paper_id string | arxiv_id string | title string | markdown dict | reviews list | scores dict | metadata dict | meta_review dict | decision dict |
|---|---|---|---|---|---|---|---|---|
k8KsI84Ds7 | 2406.01940v2 | Process-Driven Autoformalization in Lean 4 | {
"content": "## Abstract\n\nAbstract Autoformalization, the conversion of natural language mathematics into formal languages, offers significant potential for advancing mathematical reasoning. However, existing efforts are limited to formal languages with substantial online corpora and struggle to keep pace with rap... | [
{
"id": "2XQvXPyqy6",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "The paper presents a new dataset FormL4 for evaluating autoformalization in Lean4. The paper also proposes a framework called PDA to improve autoformalization cap... | {
"rating": "1;5;5;8",
"rating_avg": 4.75,
"confidence": "4;3;3;2",
"confidence_avg": 3,
"soundness": "1;2;3;2",
"soundness_avg": 2,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "1;3;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.346767"
} | {
"id": "b9J1ZiHkLn",
"metareview": "The paper takes on a unique and novel angle for advancing reasoning abilities of LLMs, by presenting a new dataset FormL4 for evaluating autoformalization in Lean4, a language with rather limited presence in training corpora. In order to improve autoformalization ability, the au... | {
"decision": "Reject"
} |
k9QklPhLCs | 2405.17506v2 | Subspace Node Pruning | {
"content": "## Abstract\n\nAbstract Efficiency of neural network inference is undeniably important in a time where commercial use of AI models increases daily.\nNode pruning is the art of removing computational units such as neurons, filters, attention heads, or even entire layers to significantly reduce inference ... | [
{
"id": "BUUFhdSFea",
"initial_rating": 3,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper proposes a post-training pruning procedure for neural networks. This procedure proceeds layer-wise, removing nodes whose activity is well explained by ... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "4;4;5;4",
"confidence_avg": 4.25,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "1;2;2;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.347930"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
kBcpBdGoKc | 2312.00343v6 | OpenStereo: A Comprehensive Benchmark for Stereo Matching and Strong Baseline | {
"content": "## Abstract\n\nAbstract Stereo matching aims to estimate the disparity between matching pixels in a stereo image pair, which is important to robotics, autonomous driving, and other computer vision tasks. Despite the development of numerous impressive methods in recent years, determining the most suitabl... | [
{
"id": "1F2yYaFHJr",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper proposes a framework for implementing different stereo models and evaluating them in a consistent manner. As part of this effort, the authors implemente... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "5;4;4;4",
"confidence_avg": 4.25,
"soundness": "2;3;3;2",
"soundness_avg": 2.5,
"contribution": "1;2;3;2",
"contribution_avg": 2,
"presentation": "2;3;3;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.348666"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
kDakBhOaBV | 2306.13840v3 | Beyond Scale: The Diversity Coefficient as a Data Quality Metric for Variability in Natural Language Data | {
"content": "## Abstract\n\nAbstract Current trends in pre-training Large Language Models (LLMs) primarily focus on the scaling of model and dataset size.\nWhile the quality of pre-training data is considered an important factor for training powerful LLMs, it remains a nebulous concept that has not been rigorously c... | [
{
"id": "RhozqvLvJL",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "The authors present a measure of text corpus diversity called \"diversity coefficient\" based on the cosine distance between Task2Vec vectors constructed from ran... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "4;3;2;4",
"confidence_avg": 3.25,
"soundness": "2;1;2;3",
"soundness_avg": 2,
"contribution": "2;1;2;3",
"contribution_avg": 2,
"presentation": "2;2;1;2",
"presentation_avg": 1.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.349443"
} | {
"id": "2USkBpfALY",
"metareview": "This paper presents a metric to quantify the diversity of a text corpus with the aim to correlate it with quality of a language model trained on such a corpus. The relies on an embedding model called Task2Vec built on top of GPT2, and the diversity score is determined by calcula... | {
"decision": "Reject"
} |
kFsWpSxkFz | 2407.08725v2 | MetaUrban: An Embodied AI Simulation Platform for Urban Micromobility | {
"content": "## Abstract\n\nAbstract 0 0 footnotetext: ∗ * ∗ Equal contribution. Public urban spaces like streetscapes and plazas serve residents and accommodate social life in all its vibrant variations.\nRecent advances in Robotics and Embodied AI make public urban spaces no longer exclusive to humans. Food delive... | [
{
"id": "JqhnzpZT8d",
"initial_rating": 8,
"confidence": 3,
"soundness": 4,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces MetaUrban, a simulation platform designed to support AI research for urban micromobility, focusing on autonomous systems navigating shared pu... | {
"rating": "5;6;8;8",
"rating_avg": 6.75,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "2;4;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.350462"
} | {
"id": "VxyXzZBDAP",
"metareview": "(a) Summary: The paper introduces MetaUrban, a simulation platform designed to support AI research for urban micromobility, focusing on autonomous systems navigating shared public spaces like streets and plazas. \n(b) Strengths: The reviewers all provided generally positive feed... | {
"decision": "Accept (Spotlight)"
} |
kGvXIlIVLM | 2410.09347v1 | Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment | {
"content": "## Abstract\n\nAbstract Classifier-Free Guidance (CFG) is a critical technique for enhancing the sample quality of visual generative models. However, in autoregressive (AR) multi-modal generation, CFG introduces design inconsistencies between language and visual content, contradicting the design philoso... | [
{
"id": "PHTuAzTuU7",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "To bridge the gap between autoregressive language and visual generation, this work targets at removing the need of CFG in AR image generation, therefore unifying ... | {
"rating": "5;6;6;6;8;8",
"rating_avg": 6.5,
"confidence": "4;4;3;3;3;4",
"confidence_avg": 3.5,
"soundness": "4;3;3;3;3;3",
"soundness_avg": 3.1666666666666665,
"contribution": "3;3;3;3;3;4",
"contribution_avg": 3.1666666666666665,
"presentation": "3;3;2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.351368"
} | {
"id": "Fyc9BEPT0U",
"metareview": "CFG improves visual generative models but creates inconsistencies in autoregressive (AR) multi-modal generation. To address this, this paper proposes Condition Contrastive Alignment (CCA), which directly fine-tunes pre-trained models to achieve guidance-free AR visual generation... | {
"decision": "Accept (Oral)"
} |
kIqA447T5c | 2410.01796v1 | Bellman Diffusion: Generative Modeling as Learning a Linear Operator in the Distribution Space | {
"content": "## Abstract\n\nAbstract Deep Generative Models (DGMs), including Energy-Based Models (EBMs) and Score-based Generative Models (SGMs), have advanced high-fidelity data generation and complex continuous distribution approximation. However, their application in Markov Decision Processes (MDPs), particularl... | [
{
"id": "ulA5S1Dw6o",
"initial_rating": 3,
"confidence": 3,
"soundness": 1,
"contribution": 2,
"presentation": 1,
"summary": "The authors propose a new genarative model, Bellman diffusion, that only depends on the density and its derivative. This choice enables learning return distributi... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "3;4;3;2",
"confidence_avg": 3,
"soundness": "1;2;3;4",
"soundness_avg": 2.5,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "1;2;2;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.352172"
} | {
"id": "N6O82gd19C",
"metareview": "This paper introduces Bellman Diffusion, which exploits advanced generative models for distributional reinforcement learning with linearity in MDPs. The method is well-motivated to fill the gap in distributional RL, and derived with the practical objectives. \n\nHowever, the maj... | {
"decision": "Reject"
} |
kJ5H7oGT2M | 2406.03386v2 | Learning Long Range Dependencies on Graphs via Random Walks | {
"content": "## Abstract\n\nAbstract Message-passing graph neural networks (GNNs) excel at capturing local relationships but struggle with long-range dependencies in graphs.\nIn contrast, graph transformers (GTs) enable global information exchange but often oversimplify the graph structure by representing graphs as ... | [
{
"id": "oNigYQC1Iu",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper addresses the challenge of capturing long-range dependencies in GNNs by combining random walks with local message passing. The authors utilize State Sp... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.353855"
} | {
"id": "SMYx8Qw4D2",
"metareview": "The paper introduces NeuralWalker, a random-walk-based graph representation method. NeuralWalker models graphs as sequences derived from random walks, encoding them through sequence layers, and integrates these encodings with local and global message-passing mechanisms. The auth... | {
"decision": "Accept (Poster)"
} |
kKILfPkhSz | 2407.00132v2 | ShortcutsBench: A Large-Scale Real-world Benchmark for API-based Agents | {
"content": "## Abstract\n\nAbstract Recent advancements in integrating large language models (LLMs) with application programming interfaces (APIs) have gained significant interest in both academia and industry. These API-based agents, leveraging the strong autonomy and planning capabilities of LLMs, can efficiently... | [
{
"id": "QM0LMeN0DB",
"initial_rating": 5,
"confidence": 4,
"soundness": 4,
"contribution": 3,
"presentation": 4,
"summary": "This paper introduces a comprehensive, large-scale benchmark designed to evaluate API-based agents in real-world scenarios. It addresses the shortcomings of exist... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "3;4;4;3",
"confidence_avg": 3.5,
"soundness": "2;3;4;2",
"soundness_avg": 2.75,
"contribution": "1;3;3;2",
"contribution_avg": 2.25,
"presentation": "1;1;4;2",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.354737"
} | {
"id": "pYLo1aeFHt",
"metareview": "The paper \"ShortcutsBench: A Large-Scale Real-world Benchmark for API-based Agents\" introduces a novel benchmark dataset, ShortcutsBench, designed to evaluate the capabilities of API-based agents in handling real-world complex tasks. The benchmark leverages real APIs from Appl... | {
"decision": "Accept (Poster)"
} |
kMT8ujhYbA | 2410.09114v2 | Catastrophic Cyber Capabilities Benchmark (3CB): Robustly Evaluating LLM Agent Cyber Offense Capabilities | {
"content": "## Abstract\n\nAbstract LLM agents have the potential to revolutionize defensive cyber operations, but their offensive capabilities are not yet fully understood. To prepare for emerging threats, model developers and governments are evaluating the cyber capabilities of foundation models. However, these a... | [
{
"id": "7X9dwaTOOY",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 1,
"presentation": 3,
"summary": "This paper proposes a framework and benchmark of 15 challenges that allows a LLM to conduct attacks. They evaluate a number of LLMs in this framework.",
"stre... | {
"rating": "3;5;6",
"rating_avg": 4.666666666666667,
"confidence": "4;4;4",
"confidence_avg": 4,
"soundness": "3;2;2",
"soundness_avg": 2.3333333333333335,
"contribution": "1;2;3",
"contribution_avg": 2,
"presentation": "3;2;3",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.355466"
} | {
"id": "LyHGMpOJSj",
"metareview": "While this work presents a wide range of benchmark tasks for the domain of cybersecurity, it is not clear, in its current state, how these evaluations provide information that is not currently available in other benchmarks, such as CyberSecEval 2. The authors mention these other... | {
"decision": "Reject"
} |
kNHVViEPWK | 2410.13826v2 | Unearthing Skill-level Insights for Understanding Trade-offs of Foundation Models | {
"content": "## Abstract\n\nAbstract With models getting stronger, evaluations have grown more complex, testing multiple skills in one benchmark and even in the same instance at once. However, skill-wise performance is obscured when inspecting aggregate accuracy, under-utilizing the rich signal modern benchmarks con... | [
{
"id": "otSfL4x958",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper uses LLMs to annotate benchmark instances with skills required to solve them. The authors validate their annotation in multiple different (mostly autom... | {
"rating": "5;5;8",
"rating_avg": 6,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "3;3;3",
"soundness_avg": 3,
"contribution": "3;2;3",
"contribution_avg": 2.6666666666666665,
"presentation": "2;3;3",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.356182"
} | {
"id": "OtU8651oa5",
"metareview": "(a) The paper proposes an automatic approach to evaluate LLMs by analyzing skills through model-generated rationales, rather than relying solely on aggregate accuracy. The method extracts skills from rationales for 46k instances across 12 benchmarks, creating \"skill-slices\" th... | {
"decision": "Accept (Poster)"
} |
kPC83HK4br | 2407.06141v1 | CHAMP: Conformalized 3D Human Multi-Hypothesis Pose Estimators | {
"content": "## Abstract\n\nAbstract We introduce CHAMP, a novel method for learning sequence-to-sequence, multi-hypothesis 3D human poses from 2D keypoints by leveraging a conditional distribution with a diffusion model. To predict a single output 3D pose sequence, we generate and aggregate multiple 3D pose hypothe... | [
{
"id": "VThp8lQtBj",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "CHAMP presents a conformalized multi-hypothesis pose estimation pipeline. Based on diffusion model, the multiple pose hypothesis are refined by a differentiable c... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "5;5;4;4",
"confidence_avg": 4.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;2;3;4",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.356943"
} | {
"id": "9nzvKaqDr5",
"metareview": "This paper addresses the task of 2D-to-3D lifting of temporal human pose sequences through a diffusion-based multi-hypothesis approach. It introduces Conformal Prediction (CP) to select more plausible hypotheses and aggregate them. The method learns a conformity score function ... | {
"decision": "Accept (Poster)"
} |
kPlePgo1Nw | 2405.15840v1 | Learning the Language of Protein Structure | {
"content": "## Abstract\n\nAbstract Representation learning and de novo generation of proteins are pivotal computational biology tasks.\nWhilst natural language processing (NLP) techniques have proven highly effective for protein sequence modelling, structure modelling presents a complex challenge, primarily due to... | [
{
"id": "wpwRYistj5",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The work proposes to use an FSQ-based auto-encoder to tokenize 3D protein structures at the residue level. They use a GNN encoder, IPA decoder, and a length compr... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;5;5;3",
"confidence_avg": 4.25,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "1;2;3;3",
"contribution_avg": 2.25,
"presentation": "2;3;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.357665"
} | {
"id": "N583qhj1qk",
"metareview": "This submission ended up with mixed reviews. While initially all reviewers except one were leaning towards recommending to reject this paper, the lively discussion with the authors led to a more split situation with two reviewers in favor of accepting and two reviewers against.\... | {
"decision": "Reject"
} |
kQNlIHlM6m | 2410.01324v1 | Fair Class-Incremental Learning using Sample Weighting | {
"content": "## Abstract\n\nAbstract Model fairness is becoming important in class-incremental learning for Trustworthy AI. While accuracy has been a central focus in class-incremental learning, fairness has been relatively understudied. However, naïvely using all the samples of the current task for training results... | [
{
"id": "EWmLytSiQu",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces Fairness-aware Sample Weighting (FSW), a novel approach for addressing unfair catastrophic forgetting in class-incremental learning. In class... | {
"rating": "3;3;6;6",
"rating_avg": 4.5,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;2;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.358434"
} | {
"id": "HDKPenAO1O",
"metareview": "This paper aims to address unfair catastrophic forgetting in class-incremental learning where the model learns new classes over time without forgetting the previous ones. The paper proposes a fairness-aware sample weighting algorithm which adjusts the training weights of current... | {
"decision": "Reject"
} |
kRBQwlkFSP | 2410.03463v1 | Diffusion State-Guided Projected Gradient for Inverse Problems | {
"content": "## Abstract\n\nAbstract Recent advancements in diffusion models have been effective in learning data priors for solving inverse problems. They leverage diffusion sampling steps for inducing a data prior while using a measurement guidance gradient at each step to impose data consistency. For general inve... | [
{
"id": "ecS2DKI0Z3",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper present a novel method for adding an add-on of projected gradient during solving inverse problems with (latent) diffusion models. The method can be app... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "4;4;3;4",
"confidence_avg": 3.75,
"soundness": "2;3;3;4",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.359256"
} | {
"id": "a7zYTl9Mot",
"metareview": "The starting point of the paper is very reasonable. Specifically, the authors aim to relieve the issue that previous methods fail to preserve the generation process on the data manifold defined by the diffusion prior when solving the inverse problem. The authors empirically find... | {
"decision": "Accept (Poster)"
} |
kS27PPs3yR | 2410.10289v1 | Fine-grained Abnormality Prompt Learning for Zero-shot Anomaly Detection | {
"content": "## Abstract\n\nAbstract Current zero-shot anomaly detection (ZSAD) methods show remarkable success in prompting large pre-trained vision-language models to detect anomalies in a target dataset without using any dataset-specific training or demonstration.\nHowever, these methods are often focused on craf... | [
{
"id": "YV5pCRsIgn",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 4,
"summary": "The paper introduces a novel framework for zero-shot anomaly detection (ZSAD) that focuses on learning fine-grained abnormality prompts without requiring detailed... | {
"rating": "5;5;5;5",
"rating_avg": 5,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "3;3;2;3",
"soundness_avg": 2.75,
"contribution": "2;2;3;2",
"contribution_avg": 2.25,
"presentation": "4;3;2;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.360211"
} | {
"id": "z3dSaFQyog",
"metareview": "The paper is well-written and the experimental results are strong and the figures are well-drawn. What reviewers are most concerned about is the novelty of this paper, they recognized that the ideas if this paper come from the prompt tuning and prototype learning, like CoCoOp. T... | {
"decision": "Reject"
} |
kSdWcw5mkp | 2405.19237v1 | ConceptPrune: Concept Editing in Diffusion Models via Skilled Neuron Pruning | {
"content": "## Abstract\n\nAbstract While large-scale text-to-image diffusion models have demonstrated impressive image-generation capabilities, there are significant concerns about their potential misuse for generating unsafe content, violating copyright, and perpetuating societal biases. Recently, the text-to-ima... | [
{
"id": "hbyMjBBmG9",
"initial_rating": 3,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The paper introduces ConceptPrune, a training-free method for concept editing in pre-trained diffusion models, specifically latent diffusion models like Stable Di... | {
"rating": "3;3;6;6",
"rating_avg": 4.5,
"confidence": "5;5;4;5",
"confidence_avg": 4.75,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.361064"
} | {
"id": "UEfzuSodg0",
"metareview": "The authors provided a detailed response that most reviewers found useful, as it partially or mostly addressed their concerns. Reviewer uDZE, who gave the lowest rating, did not actively participate in the discussion phase. The remaining reviewers maintained a positive stance on... | {
"decision": "Accept (Poster)"
} |
kSvoX0xdlO | 2311.06835v4 | Open-Set Graph Anomaly Detection via Normal Structure Regularisation | {
"content": "## Abstract\n\nAbstract This paper considers an important Graph Anomaly Detection (GAD) task, namely open-set GAD, which aims to train a detection model using a small number of normal and anomaly nodes (referred to as seen anomalies ) to detect both seen anomalies and unseen anomalies ( i.e. , anomalies... | [
{
"id": "CPpO6PJiMz",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper tackles the challenge of open-set graph anomaly detection, where the detection model is trained to detect both known and previously unseen anomalies. T... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "3;2;2;2",
"soundness_avg": 2.25,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "1;1;3;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.361769"
} | {
"id": "zcjbcWvy7a",
"metareview": "This paper focuses on training a detection model using a small number of normal and anomaly nodes to detect both seen anomalies and unseen anomalies and enhancing the generalization for both seen and unseen anomaly nodes. It introduces a regularizer to enforce the learning of co... | {
"decision": "Accept (Poster)"
} |
kTXChtaaNO | 2410.01208v2 | StringLLM: Understanding the String Processing Capability of Large Language Models | {
"content": "## Abstract\n\nAbstract String processing, which mainly involves the analysis and manipulation of strings, is a fundamental component of modern computing. Despite the significant advancements of large language models (LLMs) in various natural language processing (NLP) tasks, their capability in string p... | [
{
"id": "yH2PjHZjY2",
"initial_rating": 8,
"confidence": 4,
"soundness": 4,
"contribution": 4,
"presentation": 3,
"summary": "First comprehensive study of string processing capabilities of LLMs. Aside from just evaluating, authors also proposed specific solutions (via fine-tuning) to imp... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "2;2;2;4",
"soundness_avg": 2.5,
"contribution": "1;2;3;4",
"contribution_avg": 2.5,
"presentation": "3;3;2;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.362859"
} | {
"id": "X445MFgGn1",
"metareview": "This paper introduces StringLLM, a method for generating datasets to benchmark string processing capabilities of LLMs. Using StringLLM, the authors create StringBench, a suite of string manipulation tasks. Evaluation shows that LLM performance on StringBench tasks trails humans.... | {
"decision": "Accept (Poster)"
} |
kUH1yPMAn7 | 2408.17003v3 | Safety Layers in Aligned Large Language Models: The Key to LLM Security | {
"content": "## Abstract\n\nAbstract Aligned LLMs are secure, capable of recognizing and refusing to answer malicious questions. However, the role of internal parameters in maintaining such security is not well understood yet, further these models can be vulnerable to security degradation when fine-tuned with non-ma... | [
{
"id": "iEnEig96sG",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper identifies specific \"safety layers\" within language models that play a critical role in differentiating between malicious and normal queries. These s... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;3;3;2",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.363602"
} | {
"id": "lpmSOEyYel",
"metareview": "This paper study the ``safety layers'' in models. Motivated by the safety layers, it further proposes a fine-tuning methods which fix the gradient of safety layers during fine-tuning to improve the safety. Before rebuttal, some reviewers have concerns about the definition of bac... | {
"decision": "Accept (Poster)"
} |
kUsXwE98Cs | 2410.21259v2 | AutoBench-V: Can Large Vision-Language Models Benchmark Themselves? | {
"content": "## Abstract\n\nAbstract Large Vision-Language Models (LVLMs) have become essential for advancing the integration of visual and linguistic information, facilitating a wide range of complex applications and tasks. However, the evaluation of LVLMs presents significant challenges as the evaluation benchmark... | [
{
"id": "oIdXPnmAeq",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "The paper presents a novel automated framework for evaluating Large Vision-Language Models and conducts a detailed experimental analysis on seven popular LVLMs.",... | {
"rating": "1;3;5;6",
"rating_avg": 3.75,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;1;3;3",
"soundness_avg": 2.25,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "1;4;3;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.364367"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
kVlfYvIqaK | 2405.13937v7 | Node-Time Conditional Prompt Learning in Dynamic Graphs | {
"content": "## Abstract\n\nAbstract Dynamic graphs capture evolving interactions between entities, such as in social networks, online learning platforms, and crowdsourcing projects. For dynamic graph modeling, dynamic graph neural networks (DGNNs) have emerged as a mainstream technique. However, they are generally ... | [
{
"id": "3ZfNPufSjr",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper studies the problem of adapting pre-trained dynamic GNNs. The authors propose DYGPROMPT, a dual prompt mechanism with node and time prompts, supported ... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.365122"
} | {
"id": "i8tstLRmja",
"metareview": "The paper proposes DYGPROMPT, a dual prompt mechanism with node and time prompts for pre-training dynamic GNNs. The problem is interesting and popular. The proposed model is reasonable. The experiment results show that the proposed method outperforms baseline methods. The paper ... | {
"decision": "Accept (Poster)"
} |
kWGHZuW5yJ | 2405.18842v2 | Enhancing Descriptive Image Quality Assessment with a Large-scale Multi-modal Dataset | {
"content": "## Abstract\n\nAbstract With the rapid advancement of Vision Language Models (VLMs), VLM-based Image Quality Assessment (IQA) seeks to describe image quality linguistically to align with human expression and capture the multifaceted nature of IQA tasks.\nHowever, current methods are still far from pract... | [
{
"id": "fG4TS8eqQx",
"initial_rating": 3,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "In this paper, a large-scale EDQA-495K dataset is proposed to fine-tune VLMs as multi-functional image quality assessors that can handle both quality assessment a... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "5;3;5;5",
"confidence_avg": 4.5,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.366218"
} | {
"id": "s8GkC3mkIL",
"metareview": "This paper proposes a multi-functional VLM-based IQA method. The proposed EDQA model can handle both the image quality assessment and comparison tasks, and full-reference and non-reference scenarios. The dataset is implemented with 35 types of synthetic distortions.\n\nHowever, ... | {
"decision": "Reject"
} |
kWRKNDU6uN | 2410.13855v1 | Diffusing States and Matching Scores: A New Framework for Imitation Learning | {
"content": "## Abstract\n\nAbstract Adversarial Imitation Learning is traditionally framed as a two-player zero-sum game between a learner and an adversarially chosen cost function, and can therefore be thought of as the sequential generalization of a Generative Adversarial Network (GAN).\nHowever, in recent years,... | [
{
"id": "ayZLcFU13l",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This work focuses on imitation learning from observations through the introduction of SMILING (Score-Matching Imitation LearnING). SMILING proposes to match the s... | {
"rating": "3;5;5;6;8",
"rating_avg": 5.4,
"confidence": "4;2;4;4;3",
"confidence_avg": 3.4,
"soundness": "3;3;3;3;4",
"soundness_avg": 3.2,
"contribution": "2;3;3;3;3",
"contribution_avg": 2.8,
"presentation": "3;2;3;3;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.367071"
} | {
"id": "FXspl4clt6",
"metareview": "All the reviewers unanimously recommend accepting this paper, given its clarity and theoretical and empirical results. Also, the author's rebuttal has sufficiently addressed some concerns and questions raised by the reviewers, such as comparisons to recent works, the importance ... | {
"decision": "Accept (Poster)"
} |
kYwTmlq6Vn | 2410.20542v1 | PaPaGei: Open Foundation Models for Optical Physiological Signals | {
"content": "## Abstract\n\nAbstract Photoplethysmography (PPG) is the most widely used non-invasive technique for monitoring biosignals and cardiovascular health, with applications in both clinical settings and consumer health through wearable devices. Current machine learning models trained on PPG signals are most... | [
{
"id": "ybWg2WKWfQ",
"initial_rating": 8,
"confidence": 4,
"soundness": 4,
"contribution": 3,
"presentation": 4,
"summary": "This paper proposes two open foundation models for PPG signals, PAPAGEI-P (patient-aware) and PAPAGEI-S (morphology-aware). These models mainly differ in their tr... | {
"rating": "3;6;6;8",
"rating_avg": 5.75,
"confidence": "4;4;5;4",
"confidence_avg": 4.25,
"soundness": "3;2;3;4",
"soundness_avg": 3,
"contribution": "1;3;2;3",
"contribution_avg": 2.25,
"presentation": "2;4;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.368026"
} | {
"id": "tzvmbylRDY",
"metareview": "This paper was reviewed by four experts in the field and received 8, 1, 8, 8 as the final ratings. The reviewers acknowledged that this work presents the first open-source foundation model for PPG, trained on open datasets with open-source code, the methods and results are well ... | {
"decision": "Accept (Poster)"
} |
kam84eEmub | 2411.02322v1 | LayerDAG: A Layerwise Autoregressive Diffusion Model for Directed Acyclic Graph Generation | {
"content": "## Abstract\n\nAbstract Directed acyclic graphs (DAGs) serve as crucial data representations in domains such as hardware synthesis and compiler/program optimization for computing systems. DAG generative models facilitate the creation of synthetic DAGs, which can be used for benchmarking computing system... | [
{
"id": "rtKP9H1Cw5",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper presents an autoregressive model, LayerDAG, for generating directed acyclic graphs (DAGs). A key contribution of the paper is framing the problem as a ... | {
"rating": "5;6;8;8;8",
"rating_avg": 7,
"confidence": "3;3;3;2;4",
"confidence_avg": 3,
"soundness": "2;3;4;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3;3",
"contribution_avg": 2.8,
"presentation": "3;3;4;3;4",
"presentation_avg": 3.4
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.368849"
} | {
"id": "0ILeITM20N",
"metareview": "This paper introduces LayerDAG, a novel approach for generating directed acyclic graphs (DAGs) that combines autoregressive generation with diffusion models. LayerDAG leverages the layered structure of DAGs, induced by their topological ordering, to model both directional and lo... | {
"decision": "Accept (Spotlight)"
} |
kaqrwQ96xW | 2410.05664v1 | Holistic Unlearning Benchmark: A Multi-Faceted Evaluation for Text-to-Image Diffusion Model Unlearning | {
"content": "## Abstract\n\nAbstract As text-to-image diffusion models become advanced enough for commercial applications, there is also increasing concern about their potential for malicious and harmful use. Model unlearning has been proposed to mitigate the concerns by removing undesired and potentially harmful in... | [
{
"id": "1l3CxjZvG0",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The Holistic Unlearning Benchmark provides a comprehensive evaluation framework for text-to-image diffusion model unlearning. The benchmark assesses unlearning me... | {
"rating": "5;5;5",
"rating_avg": 5,
"confidence": "3;3;3",
"confidence_avg": 3,
"soundness": "4;2;3",
"soundness_avg": 3,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "4;3;3",
"presentation_avg": 3.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.369608"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
kat8uANDlU | 2411.04899v1 | Sampling-guided Heterogeneous Graph Neural Network with Temporal Smoothing for Scalable Longitudinal Data Imputation | {
"content": "## Abstract\n\nAbstract In this paper, we propose a novel framework, the Sampling-guided Heterogeneous Graph Neural Network (SHT-GNN), to effectively tackle the challenge of missing data imputation in longitudinal studies. Unlike traditional methods, which often require extensive preprocessing to handle... | [
{
"id": "NG8gATHAaz",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes Sampling-guided Heterogeneous Graph Neural Network (SHT-GNN) for Longitudinal data imputation. SHT-GNN treats both observations and covariates ... | {
"rating": "5;5;5;6;6",
"rating_avg": 5.4,
"confidence": "4;4;3;2;3",
"confidence_avg": 3.2,
"soundness": "2;3;2;3;3",
"soundness_avg": 2.6,
"contribution": "2;3;2;3;3",
"contribution_avg": 2.6,
"presentation": "3;2;2;3;3",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.370273"
} | {
"id": "SEgJBlFLbE",
"metareview": "This work proposed a GNN model for longitudinal data imputation, where the observations, covariates, temporal and subject information are jointly considered in a heterogeneous graph. The reviewers agree that the paper addressed an important problem in dealing with longitudinal d... | {
"decision": "Reject"
} |
kbSU5bwoRv | 2407.07728v4 | SaMoye: Zero-shot Singing Voice Conversion Model Based on Feature Disentanglement and Enhancement | {
"content": "## Abstract\n\nAbstract Singing voice conversion (SVC) aims to convert a singer’s voice to another singer’s from a reference audio while keeping the original semantics. However, existing SVC methods can hardly perform zero-shot due to incomplete feature disentanglement or dependence on the speaker look-... | [
{
"id": "uCz6fjstYA",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "The paper presents SaMoye, a method for singing-voice conversion in the zero-shot setting. The method extends existing approaches for SVC in a few different ways:... | {
"rating": "3;3;3;3;3",
"rating_avg": 3,
"confidence": "4;5;5;5;3",
"confidence_avg": 4.4,
"soundness": "2;2;2;2;3",
"soundness_avg": 2.2,
"contribution": "2;2;1;1;2",
"contribution_avg": 1.6,
"presentation": "2;2;1;1;2",
"presentation_avg": 1.6
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.371058"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
kcAejITM7C | 2410.02892v1 | The Role of Deductive and Inductive Reasoning in Large Language Models | {
"content": "## Abstract\n\nAbstract Large Language Models (LLMs) have achieved substantial progress in artificial intelligence, particularly in reasoning tasks. However, their reliance on static prompt structures, coupled with limited dynamic reasoning capabilities, often constrains their adaptability to complex an... | [
{
"id": "rC2iJsu490",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper presents the De-In-Ductive (DID) prompt strategy, which integrates deductive and inductive reasoning to improve adaptability in large language models (... | {
"rating": "3;3;3;3;5",
"rating_avg": 3.4,
"confidence": "3;4;3;3;3",
"confidence_avg": 3.2,
"soundness": "2;2;2;2;2",
"soundness_avg": 2,
"contribution": "2;3;1;2;2",
"contribution_avg": 2,
"presentation": "3;2;2;1;3",
"presentation_avg": 2.2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.371640"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
kfsgz2Ldd7 | 2410.05646v1 | Score-Based Variational Inference for Inverse Problems | {
"content": "## Abstract\n\nAbstract Existing diffusion-based methods for inverse problems sample from the posterior using score functions and accept the generated random samples as solutions. In applications that posterior mean is preferred, we have to generate multiple samples from the posterior which is time-cons... | [
{
"id": "xM6JOiGN0u",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 1,
"presentation": 2,
"summary": "The authors propose Reverse Mean Propagation (RMP) as a new framework that leverages score-based variational inference to solve inverse problems by tracking the p... | {
"rating": "3;6;6;6;6",
"rating_avg": 5.4,
"confidence": "4;2;4;3;4",
"confidence_avg": 3.4,
"soundness": "2;3;3;3;3",
"soundness_avg": 2.8,
"contribution": "1;3;3;3;3",
"contribution_avg": 2.6,
"presentation": "2;3;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.372277"
} | {
"id": "IjsAT8gmNu",
"metareview": "This paper proposes a variational inference method for score-based approach to Bayesian inverse problems. The main contribution is a reverse mean propogation method (RMP) that directly approximates the posterior mean, which is implemented by solving multiple variational problems... | {
"decision": "Reject"
} |
kiOxNsrpQy | 2406.15156v1 | Reconsidering Faithfulness in Regular, Self-Explainable and Domain Invariant GNNs | {
"content": "## Abstract\n\nAbstract As Graph Neural Networks (GNNs) become more pervasive, it becomes paramount to build robust tools for computing explanations of their predictions.\nA key desideratum is that these explanations are faithful , i.e. , that they portray an accurate picture of the GNN’s reasoning proc... | [
{
"id": "cqpebYyKAr",
"initial_rating": 5,
"confidence": 2,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper addresses a timely and impactful topic by exploring the limitations of current metrics used to evaluate the faithfulness of explanations in graph neura... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "3;2;3;4",
"confidence_avg": 3,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;2;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.373154"
} | {
"id": "cbOOu8YZkb",
"metareview": "This is an important and timely contribution to faithfulness in GNNs. The Reviewers are unanimously in support of accepting the paper post-rebuttal, and I believe it will spark important discussions at the conference. A clear accept.",
"additional_comments": "Initially the maj... | {
"decision": "Accept (Poster)"
} |
kjmLabjSE2 | 2410.01068v1 | Convergent Privacy Loss of Noisy-SGD without Convexity and Smoothness | {
"content": "## Abstract\n\nAbstract We study the Differential Privacy (DP) guarantee of hidden-state Noisy-SGD algorithms over a bounded domain. Standard privacy analysis for Noisy-SGD assumes all internal states are revealed, which leads to a divergent Rényi DP bound with respect to the number of iterations. Ye & ... | [
{
"id": "Uq5B2bLXVa",
"initial_rating": 8,
"confidence": 3,
"soundness": 4,
"contribution": 4,
"presentation": 4,
"summary": "The paper refines the hidden-state privacy analysis of Noisy-SGD algorithm under various loss assumptions from strong-convex to non-convex and from Lipschitz smoo... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "4;4;3;4",
"soundness_avg": 3.75,
"contribution": "3;2;2;4",
"contribution_avg": 2.75,
"presentation": "3;3;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.374380"
} | {
"id": "b6BYEpfqPl",
"metareview": "The paper got sufficient support from the reviewers (with a few of them have historically worked on similar problems). However, there were a few major concerns, a) The assumption on Holder continuous gradient is unclear how operational it is for natural optimization tasks, and b... | {
"decision": "Accept (Poster)"
} |
kkE7jlqKae | 2406.09398v1 | Real-Time Deepfake Detection in the Real World | {
"content": "## Abstract\n\nAbstract Recent improvements in generative AI made synthesizing fake images easy; as they can be used to cause harm, it is crucial to develop accurate techniques to identify them. This paper introduces \"Locally Aware Deepfake Detection Algorithm\" ( LaDeDa ), that accepts a single 9 × 9 ... | [
{
"id": "6WHJFdgMh4",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper introduces a novel deepfake detection algorithm named Locally Aware Deepfake Detection Algorithm (LaDeDa), which processes individual 9x9 image patches ... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "4;2;3;5",
"confidence_avg": 3.5,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "3;2;2;3",
"contribution_avg": 2.5,
"presentation": "3;3;3;2",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.375242"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
kkGIbmpCHU | 2410.21523v1 | Diffusion-Nested Auto-Regressive Synthesis of Heterogeneous Tabular Data | {
"content": "## Abstract\n\nAbstract Autoregressive models are predominant in natural language generation, while their application in tabular data remains underexplored. We posit that this can be attributed to two factors: 1) tabular data contains heterogeneous data type, while the autoregressive model is primarily ... | [
{
"id": "5qxG3J7wvc",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "The authors develop a generative modeling method for tabular data. Their method learns a conditional generative model that takes masked input and predicts the mi... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "3;3;4;4",
"confidence_avg": 3.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "2;3;3;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.375940"
} | {
"id": "9TwFQCPkEI",
"metareview": "The paper proposes a generative model for tabular data called Diffusion-nested Autoregressive model (TabDAR), which combines masked bi-directional attention transformers with diffusion models to handle mixed numerical and categorical data and supports arbitrary generation orders... | {
"decision": "Reject"
} |
kmgrlG9TR0 | 2410.09893v1 | RMB: Comprehensively benchmarking reward models in LLM alignment | {
"content": "## Abstract\n\nAbstract Reward models (RMs) guide the alignment of large language models (LLMs), steering them toward behaviors preferred by humans. Evaluating RMs is the key to better aligning LLMs.\nHowever, the current evaluation of RMs may not directly correspond to their alignment performance due t... | [
{
"id": "S0Q5FFExeB",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper presents RMB, a benchmark evaluating reward models (RMs) across 49 real-world scenarios, improving alignment assessments for LLMs. RMB reveals generaliz... | {
"rating": "5;6;6",
"rating_avg": 5.666666666666667,
"confidence": "3;4;4",
"confidence_avg": 3.6666666666666665,
"soundness": "3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3",
"contribution_avg": 2.6666666666666665,
"presentation": "2;4;2",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.376813"
} | {
"id": "LXBYgvcKG8",
"metareview": "**(a) Summarize the scientific claims and findings**\n\nThe paper introduces RMB (Reward Model Benchmark), a comprehensive benchmark designed to evaluate reward models (RMs) used in aligning large language models (LLMs). RMB includes 49 fine-grained real-world scenarios with two... | {
"decision": "Accept (Poster)"
} |
kp3Trt4uvf | 2408.06958v2 | AuToMATo: An Out-Of-The-Box Persistence-Based Clustering Algorithm | {
"content": "## Abstract\n\nAbstract We present AuToMATo, a novel clustering algorithm based on persistent homology. While AuToMATo is not parameter-free per se, we provide default choices for its parameters that make it into an out-of-the-box clustering algorithm that performs well across the board. AuToMATo combin... | [
{
"id": "JAwoBxrBc3",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper considers and extends the ToMATo clustering algorithm, which finds clusters based on the persistent homology of a density estimate of the data. This pa... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "4;5;3;3",
"confidence_avg": 3.75,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "1;1;3;2",
"contribution_avg": 1.75,
"presentation": "2;2;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.377654"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
kpL66Mvd2a | 2407.01476v2 | Tree Search for Language Model Agents | {
"content": "## Abstract\n\nAbstract Autonomous agents powered by language models (LMs) have demonstrated promise in their ability to perform decision-making tasks such as web automation. However, a key limitation remains: LMs, primarily optimized for natural language understanding and generation, struggle with mult... | [
{
"id": "l7nkrKm4xP",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper introduces a tree search algorithm for improving the performance of language model agents. By implementing a best-first search approach that operates i... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "5;4;3;4",
"confidence_avg": 4,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;2;3;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.378404"
} | {
"id": "jNniys0Ufk",
"metareview": "In this paper, the authors propose an inference-time tree search algorithm for LLM agents to perform web exploration and planning tasks. Reviewers noted that the experiments demonstrate the effectiveness of the approach and its compatibility with various LLM models. However, the... | {
"decision": "Reject"
} |
kpPjV2krAa | 2409.19111v2 | FUSION IS ALL YOU NEED : FACE FUSION FOR CUSTOMIZED IDENTITY-PRESERVING IMAGE SYNTHESIS | {
"content": "## Abstract\n\nAbstract Text-to-image (T2I) models have significantly advanced the development of artificial intelligence, enabling the generation of high-quality images in diverse contexts based on specific text prompts. However, existing T2I-based methods often struggle to accurately reproduce the app... | [
{
"id": "xtHLc2Pr06",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper improves text-to-image generation by preserving individual identities in generated images. Unlike previous methods, it directly incorporates face image... | {
"rating": "3;3;3;6",
"rating_avg": 3.75,
"confidence": "5;4;4;2",
"confidence_avg": 3.75,
"soundness": "3;2;2;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;2;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.379099"
} | {
"id": "It0wWQZ89S",
"metareview": "The paper proposes an identity-preserving text-to-image generation method by modifying the UNet architecture in Stable Diffusion. Instead of using a separate image encoder, the method directly incorporates reference facial images into the cross-attention layers, aiming to improv... | {
"decision": "Reject"
} |
kpnW12Lm9p | 2403.13838v1 | Circuit Transformer: A Transformer That Preserves Logical Equivalence | {
"content": "## Abstract\n\nAbstract. Language, a prominent human ability to express through sequential symbols, has been computationally mastered by recent advances of large language models (LLMs). By predicting the next word recurrently with huge neural models, LLMs have shown unprecedented capabilities in underst... | [
{
"id": "l7rptZRmBZ",
"initial_rating": 8,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 2,
"summary": "Summary:\nThis paper introduces an approach called \"Circuit Transformer\" for generating logically equivalent circuits while preserving strict equivalence constr... | {
"rating": "3;6;6",
"rating_avg": 5,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;3;3",
"contribution_avg": 2.6666666666666665,
"presentation": "2;2;3",
"presentation_avg": 2.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.379878"
} | {
"id": "xMe3DVvLao",
"metareview": "This paper introduces “circuit” transformer which generate logical circuits. Furthermore it can ensure the generated circuits’ equivalence with the given boolean function. The main novelty lies on its proposes decoding mechanism based on a cutoff property that prevents logical e... | {
"decision": "Accept (Poster)"
} |
kpq3IIjUD3 | 2407.06053v3 | Learning local equivariant representations for quantum operators | {
"content": "## Abstract\n\nAbstract Predicting quantum operator matrices such as Hamiltonian, overlap, and density matrices in the density functional theory (DFT) framework is crucial for understanding material properties. Current methods often focus on individual operators and struggle with efficiency and scalabil... | [
{
"id": "2SLn7XVwPu",
"initial_rating": 6,
"confidence": 3,
"soundness": 4,
"contribution": 3,
"presentation": 3,
"summary": "This paper proposes a data driven method to solve KS-DFT. Instead of solving the KS system to consistency, this method puts the configuration of atoms through a c... | {
"rating": "3;6;6",
"rating_avg": 5,
"confidence": "5;3;3",
"confidence_avg": 3.6666666666666665,
"soundness": "2;3;4",
"soundness_avg": 3,
"contribution": "2;3;3",
"contribution_avg": 2.6666666666666665,
"presentation": "3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.380658"
} | {
"id": "sz3mjvk8q2",
"metareview": "The paper proposes an equivariant neural network that predicts Hamiltonian (converged Fock matrix in DFT) for molecular systems. The authors highlighted the benefit of reduced computational and memory cost by imposing a strictly local design to the architecture, together with ot... | {
"decision": "Accept (Spotlight)"
} |
kqSmedTcgb | 2407.15141v1 | Text-Augmented Multimodal LLMs for Chemical Reaction Condition Recommendation | {
"content": "## Abstract\n\nAbstract High-throughput reaction condition (RC) screening is fundamental to chemical synthesis. However, current RC screening suffers from laborious and costly trial-and-error workflows. Traditional computer-aided synthesis planning (CASP) tools fail to find suitable RCs due to data spar... | [
{
"id": "bhnvPhgYa7",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper introduces Chemma-RC, a multimodal large language model designed for recommending chemical reaction conditions. It integrates SMILES, reaction graphs, a... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;4;5;3",
"confidence_avg": 4,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "2;2;3;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.381449"
} | {
"id": "lP8jtuckoC",
"metareview": "The work introduces a multimodal LLM-based model for predicting chemical reaction conditions and yield.\n\nTwo reviewers voted for accepting the work, while two other voted for rejection. Among the strengths, reviewers (WdLV and 3FKb) noted the extensive instruction tuning datas... | {
"decision": "Reject"
} |
krJ73n4Pma | 2406.06751v1 | Complexity-Aware Deep Symbolic Regression with Robust Risk-Seeking Policy Gradients | {
"content": "## Abstract\n\nAbstract This paper proposes a novel deep symbolic regression approach to enhance the robustness and interpretability of data-driven mathematical expression discovery. Despite the success of the state-of-the-art method, DSR, it is built on recurrent neural networks, purely guided by data ... | [
{
"id": "FXkgGzc6Xv",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper presents CADSR, a novel approach to symbolic regression, addressing gradient vanishing and overfitting limitations of models like DSR by replacing RNN ... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "3;2;3;2",
"contribution_avg": 2.5,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.382095"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
kvLenbZZgg | 2407.07810v2 | Transformer Block Coupling and its Correlation with Generalization in LLMs | {
"content": "## Abstract\n\nAbstract Large Language Models (LLMs) have made significant strides in natural language\nprocessing, and a precise understanding of the internal mechanisms driving their\nsuccess is essential. In this work, we trace the trajectories of individual tokens as they pass through transformer bl... | [
{
"id": "cupyM21bFS",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper presents an insightful exploration into the internal dynamics of Large Language Models (LLMs). The authors propose a novel framework to evaluate the co... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "3;4;2;2",
"confidence_avg": 2.75,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "1;2;2;3",
"contribution_avg": 2,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.382713"
} | {
"id": "0HRJDACDp4",
"metareview": "This work investigates the relationship between transformer block coupling and the generalization capabilities of large language models (LLMs). The authors claim that the interaction between different transformer blocks significantly influences the model's ability to generalize ... | {
"decision": "Accept (Poster)"
} |
kwY3eL3QVh | 2410.11646v1 | Feature-guided score diffusion for sampling conditional densities | {
"content": "## Abstract\n\nAbstract Score diffusion methods can learn probability densities from samples. The score of the noise-corrupted density is estimated using a deep neural network, which is then used to iteratively transport a Gaussian white noise density to a target density. Variants for conditional densit... | [
{
"id": "emivcvXUhk",
"initial_rating": 6,
"confidence": 4,
"soundness": 4,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces a novel guided score-based diffusion model that does not require any additional structural modifications. Instead, it extracts features from... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "3;4;4;3",
"confidence_avg": 3.5,
"soundness": "1;2;4;3",
"soundness_avg": 2.5,
"contribution": "1;2;3;3",
"contribution_avg": 2.25,
"presentation": "1;3;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.383433"
} | {
"id": "QpFnVgQC48",
"metareview": "The authors present a method for score-based diffusion that doesn't depend on score guidance approximations. The four reviewers agreed that the problem is very relevant to current research directions and addresses a known issue in the field. The reviews were split two positive a... | {
"decision": "Reject"
} |
kwqhn2VuG4 | 2406.08418v3 | OmniCorpus: A Unified Multimodal Corpus of 10 Billion-Level Images Interleaved with Text | {
"content": "## Abstract\n\nAbstract Image-text interleaved data, consisting of multiple images and texts arranged in a natural document format, aligns with the presentation paradigm of internet data and closely resembles human reading habits.\nRecent studies have shown that such data aids multimodal in-context lear... | [
{
"id": "r0kYxpYJha",
"initial_rating": 6,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "The paper presents OmniCorpus, a large multimodal (text and vision) and multilingual (English and Chinese) dataset containing bilions of images interleaved with t... | {
"rating": "6;8;8;8",
"rating_avg": 7.5,
"confidence": "4;5;3;3",
"confidence_avg": 3.75,
"soundness": "2;4;3;3",
"soundness_avg": 3,
"contribution": "3;4;3;3",
"contribution_avg": 3.25,
"presentation": "3;4;3;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.384076"
} | {
"id": "jKL0zuRFAC",
"metareview": "The paper proposes a very large dataset for MLLM research, including video. Its size, diversity and benchmarking form a significant contribution. Some concerns were raised about the ethics of the data collection and some aspects of the experiments, but seem well addressed. Three... | {
"decision": "Accept (Spotlight)"
} |
kxFtMHItrf | 2311.11638v2 | Reti-Diff: Illumination Degradation Image Restoration with Retinex-based Latent Diffusion Model | {
"content": "## Abstract\n\nAbstract Illumination degradation image restoration (IDIR) techniques aim to improve the visibility of degraded images and mitigate the adverse effects of deteriorated illumination. Among these algorithms, diffusion model (DM)-based methods have shown promising performance but are often b... | [
{
"id": "CAdAWaToQ6",
"initial_rating": 6,
"confidence": 5,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper proposes to use diffusion models in conjunction with a transformer model to perform image restoration of poorly lit images. The proposed method is evalu... | {
"rating": "3;6;6;8",
"rating_avg": 5.75,
"confidence": "5;5;5;5",
"confidence_avg": 5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "1;3;3;3",
"contribution_avg": 2.5,
"presentation": "2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.384768"
} | {
"id": "mC1mRaT07k",
"metareview": "(a) Summary:\nThe paper introduces Reti-Diff, a novel framework leveraging Retinex-based Latent Diffusion Models (RLDM) and Retinex-guided Transformers (RGformer) for illumination degradation image restoration (IDIR). The method estimates compact Retinex priors to enhance degrad... | {
"decision": "Accept (Spotlight)"
} |
kxnoqaisCT | 2410.05243v1 | Navigating the Digital World as Humans Do: Universal Visual Grounding for GUI Agents | {
"content": "## Abstract\n\nAbstract Multimodal large language models (MLLMs)\nare transforming the capabilities of graphical user interface (GUI) agents, facilitating their transition from controlled simulations to complex, real-world applications across various platforms.\nHowever, the effectiveness of these agent... | [
{
"id": "wr0LkoMj69",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces a vision-only approach to GUI agents that mimics how humans interact with interfaces through visual perception and pixel-level operations. M... | {
"rating": "5;6;8;8",
"rating_avg": 6.75,
"confidence": "4;5;5;4",
"confidence_avg": 4.5,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;4;4;3",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.385510"
} | {
"id": "hcEU7LEkMs",
"metareview": "This reviewers generally liked the work, which presents a vision-only approach for GUI agents, using a visual grounding model (UGround) trained on a large synthetic dataset (Web-Hybrid). The authors argue this approach is more human-like and avoids the noise and overhead of text... | {
"decision": "Accept (Oral)"
} |
kyVzYpDxHg | 2406.01552v1 | Learning equivariant tensor functions with applications to sparse vector recovery | {
"content": "## Abstract\n\nAbstract This work characterizes equivariant polynomial functions from tuples of tensor inputs to tensor outputs. Loosely motivated by physics, we focus on equivariant functions with respect to the diagonal action of the orthogonal group on tensors. We show how to extend this characteriza... | [
{
"id": "4PAGUZdppH",
"initial_rating": 5,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This work characterizes equivariant polynomial functions from tuples of tensor inputs to tensor outputs. The goal behind these characterizations is to define equi... | {
"rating": "5;5;5;8",
"rating_avg": 5.75,
"confidence": "3;3;5;4",
"confidence_avg": 3.75,
"soundness": "3;3;2;4",
"soundness_avg": 3,
"contribution": "2;2;2;4",
"contribution_avg": 2.5,
"presentation": "1;4;2;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.386651"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
kynD1UUk6q | 2410.04472v2 | Collapsed Language Models Promote Fairness | {
"content": "## Abstract\n\nAbstract To mitigate societal biases implicitly encoded in recent successful pretrained language models, a diverse array of approaches have been proposed to encourage model fairness, focusing on prompting, data augmentation, regularized fine-tuning, and more.\nDespite the development, it ... | [
{
"id": "TeODoga2Wu",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "This work explores fairness in language models from the perspective of Neural Collapse (NC), a phenomenon observed during the final stages of training neural netw... | {
"rating": "5;5;8;8",
"rating_avg": 6.5,
"confidence": "4;3;3;3",
"confidence_avg": 3.25,
"soundness": "2;2;4;3",
"soundness_avg": 2.75,
"contribution": "2;3;3;2",
"contribution_avg": 2.5,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.387787"
} | {
"id": "UznRMKiUOa",
"metareview": "(a) **summary** This paper investigates the connection between Neural Collapse (NC) -- a learning phenomenon that happens in last-layer representations and classifiers in deep networks -- and fairness in language models (LMs). The authors hypothesize that similar effects may man... | {
"decision": "Accept (Poster)"
} |
l0fn10vSyM | 2405.01924v1 | Semi-Parametric Retrieval via Binary Bag-of-Tokens Index | {
"content": "## Abstract\n\nAbstract The landscape of information retrieval has broadened from search services to a critical component in various advanced applications, where indexing efficiency, cost-effectiveness, and freshness are increasingly important yet remain less explored.\nTo address these demands, we intr... | [
{
"id": "oj7YhgxpMd",
"initial_rating": 5,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "This paper main contribution is a retrieval model where documents are represented by the presence (boolean) of tokens (i.e. a binary bag of tokens). The advantage... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;5;3;4",
"confidence_avg": 4,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"presentation": "3;4;2;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.388535"
} | {
"id": "60SpBv2GIM",
"metareview": "This paper introduces a new, efficient retrieval method, namely “semi-parametric disentangled retrieval framework (SiDR)”. The key idea is to have two encoders: (1) a VDR-based neural encoder (called parametric), and (2) a simple bag-of-words representation (called nonparametric... | {
"decision": "Accept (Poster)"
} |
l0tg0jzsdL | 2410.09724v1 | Taming Overconfidence in LLMs: Reward Calibration in RLHF | {
"content": "## Abstract\n\nAbstract Language model calibration refers to the alignment between the confidence of the model and the actual performance of its responses.\nWhile previous studies point out the overconfidence phenomenon in Large Language Models (LLMs) and show that LLMs trained with Reinforcement Learni... | [
{
"id": "uQnTkqJj1O",
"initial_rating": 6,
"confidence": 4,
"soundness": 4,
"contribution": 2,
"presentation": 4,
"summary": "In instruction-following for language models, the paper studies the task of reporting a confidence score in the helpfulness and accuracy of the model's own respon... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;4;3;4",
"confidence_avg": 3.75,
"soundness": "2;2;2;4",
"soundness_avg": 2.5,
"contribution": "2;3;2;2",
"contribution_avg": 2.25,
"presentation": "4;3;3;4",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.389295"
} | {
"id": "KPaWi1uTnJ",
"metareview": "This paper investigates overconfidence in LLMs trained with RLHF, finding that reward models are biased towards high-confidence responses regardless of accuracy. To address this, the authors propose two heuristic methods: PPO-M, which calibrates reward models, and PPO-C, which d... | {
"decision": "Accept (Poster)"
} |
l1raPjOUPA | 2410.04489v1 | Grokking at the Edge of Linear Separability | {
"content": "## Abstract\n\nAbstract We study the generalization properties of binary logistic classification in a simplified setting, for which a \"memorizing\" and \"generalizing\" solution can always be strictly defined, and elucidate empirically and analytically the mechanism underlying Grokking in its dynamics.... | [
{
"id": "RfR3Ilgffc",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper identifies a simple linear classification setting and studies when \"grokking\" occurs, under Gaussian random data in the linear asymptotic regime. The ... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "3;3;4;4",
"confidence_avg": 3.5,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;1;2;2",
"contribution_avg": 1.75,
"presentation": "2;2;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.390217"
} | {
"id": "mUUYtzLjJw",
"metareview": "This paper theoretically investigates the dynamics of binary logistic classification in a toy setup. By identifying the \"memorization\" and \"generalization\" solutions to the problem, the paper provides interesting theoretical insights into Grokking dynamics, which have been o... | {
"decision": "Reject"
} |
l2odw7OiNw | 2409.08770v3 | Increasing Both Batch Size and Learning Rate Accelerates Stochastic Gradient Descent | {
"content": "## Abstract\n\nAbstract The performance of mini-batch stochastic gradient descent (SGD) strongly depends on setting the batch size and learning rate to minimize the empirical loss in training the deep neural network.\nIn this paper, we present theoretical analyses of mini-batch SGD with four schedulers:... | [
{
"id": "sXhCtw3cqu",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper presents a convergence analysis of mini-batch stochastic gradient descent which highlights the interplay of batch size and learning rate schedules. The... | {
"rating": "1;3;3;3",
"rating_avg": 2.5,
"confidence": "5;4;3;3",
"confidence_avg": 3.75,
"soundness": "2;2;2;2",
"soundness_avg": 2,
"contribution": "1;2;1;2",
"contribution_avg": 1.5,
"presentation": "2;3;2;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.391258"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
l2zFn6TIQi | 2410.23054v1 | Controlling Language and Diffusion Models by Transporting Activations | {
"content": "## Abstract\n\nAbstract The increasing capabilities of large generative models and their ever more widespread deployment have raised concerns about their reliability, safety, and potential misuse. To address these issues, recent works have proposed to control model generation by steering model activatio... | [
{
"id": "pnKeqD72XI",
"initial_rating": 8,
"confidence": 3,
"soundness": 4,
"contribution": 4,
"presentation": 4,
"summary": "The paper introduces Activation Transport (ACT), a framework based on optimal transport theory to steer model activations and control generative model outputs. AC... | {
"rating": "6;6;6;8",
"rating_avg": 6.5,
"confidence": "3;4;4;3",
"confidence_avg": 3.5,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "3;3;3;4",
"contribution_avg": 3.25,
"presentation": "2;4;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.392119"
} | {
"id": "ekmbX0eBrX",
"metareview": "The paper addresses the challenge of activation steering for controlling generative models by learning an optimal transport map between source and target activation distributions while preserving the target distributions. The authors highlight that existing activation-steering a... | {
"decision": "Accept (Spotlight)"
} |
l3YIMopcR9 | 2406.07411v2 | VersiCode: Towards Version-controllable Code Generation | {
"content": "## Abstract\n\nAbstract Large Language Models (LLMs) have made tremendous strides in code generation, but existing research fails to account for the dynamic nature of software development, marked by frequent library updates.\nThis gap significantly limits LLMs’ deployment in realistic settings.\nIn this... | [
{
"id": "F6pcE0ybwV",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "Despite current LLMs' remarkable advances in code generation tasks, they are typically trained and evaluated using static benchmarks such as HumanEval and MBPP. H... | {
"rating": "3;5;6;6;6",
"rating_avg": 5.2,
"confidence": "4;4;4;3;4",
"confidence_avg": 3.8,
"soundness": "2;3;3;4;3",
"soundness_avg": 3,
"contribution": "2;3;3;3;2",
"contribution_avg": 2.6,
"presentation": "2;3;2;3;3",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.393140"
} | {
"id": "1ugIddi9ty",
"metareview": "This paper addresses the issue that code generated by LLMs needs to target specific library versions, a problem overlooked in previous work. While the reviewers found the topic interesting and recognized its potential, several concerns were raised. Several expert reviewers have ... | {
"decision": "Reject"
} |
l4jBHP4FPy | 2410.02675v2 | FAN: Fourier Analysis Networks | {
"content": "## Abstract\n\nAbstract Despite the remarkable success achieved by neural networks, particularly those represented by MLP and Transformer, we reveal that they exhibit potential flaws in the modeling and reasoning of periodicity, i.e., they tend to memorize the periodic data rather than genuinely underst... | [
{
"id": "jby3D0O3Ir",
"initial_rating": 1,
"confidence": 5,
"soundness": 2,
"contribution": 1,
"presentation": 2,
"summary": "The paper addresses the critical challenge of periodicity modeling in neural networks, proposing the Fourier Analysis Network (FAN) to overcome inherent limitatio... | {
"rating": "1;3;5;6",
"rating_avg": 3.75,
"confidence": "5;3;3;3",
"confidence_avg": 3.5,
"soundness": "2;1;3;3",
"soundness_avg": 2.25,
"contribution": "1;2;2;3",
"contribution_avg": 2,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.394090"
} | {
"id": "hjAJWk3olC",
"metareview": "The paper presents a Fourier Analysis Network (FAN) to mitigate the inherent limitation of MLP and transformers of being unable to model periodicity in data. The method is novel and is intuitive considering periodic nature of fourier analysis. \n\nThe paper suffers from the foll... | {
"decision": "Reject"
} |
l6QnSQizmN | 2302.02182v2 | Online Reinforcement Learning in Non-Stationary Context-Driven Environments | {
"content": "## Abstract\n\nAbstract We study online reinforcement learning (RL) in non-stationary environments, where a time-varying exogenous context process affects the environment dynamics. Online RL is challenging in such environments due to “catastrophic forgetting” (CF). The agent tends to forget prior knowle... | [
{
"id": "AcTfiMGws0",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes Locally Constrained Policy Optimization (LCPO), a policy gradient algorithm for non-stationary reinforcement learning where the context changes... | {
"rating": "3;6;6;8",
"rating_avg": 5.75,
"confidence": "3;3;4;4",
"confidence_avg": 3.5,
"soundness": "2;3;3;4",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.394959"
} | {
"id": "IyN0eaHzKX",
"metareview": "The paper proposes Locally Constrained Policy Optimization (LCPO), which is a policy gradient method for changing environments. The reviewers highlight that the problem setting is well specified and reasonable, the proposed algorithm is a straightforward extension of TRPO, and t... | {
"decision": "Accept (Spotlight)"
} |
l8zRnvD95l | 2406.04940v1 | CarbonSense: A Multimodal Dataset and Baseline for Carbon Flux Modelling | {
"content": "## Abstract\n\nAbstract Terrestrial carbon fluxes provide vital information about our biosphere’s health and its capacity to absorb anthropogenic CO 2 emissions. The importance of predicting carbon fluxes has led to the emerging field of data-driven carbon flux modelling (DDCFM), which uses statistical ... | [
{
"id": "qTr0Sdlwz9",
"initial_rating": 6,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposes a standardized dataset called CarbonSense, which is a dataset compiled from various sources. The compilation steps include fusing the data tog... | {
"rating": "6;6;6",
"rating_avg": 6,
"confidence": "5;3;5",
"confidence_avg": 4.333333333333333,
"soundness": "3;3;2",
"soundness_avg": 2.6666666666666665,
"contribution": "3;3;2",
"contribution_avg": 2.6666666666666665,
"presentation": "3;4;3",
"presentation_avg": 3.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.395998"
} | {
"id": "HHDg7QqzCI",
"metareview": "The paper presents an ML ready dataset for carbon flux modeling that includes multiple modalities (measure fluxes, satellite images, other predictors) making it a nice comprehensive dataset for use by the community. Further, they present a novel transformer model (based on the P... | {
"decision": "Accept (Poster)"
} |
l9LWx9HMl5 | 2410.08458v1 | Simultaneous Reward Distillation and Preference Learning: Get You a Language Model Who Can Do Both | {
"content": "## Abstract\n\nAbstract Reward modeling of human preferences is one of the cornerstones of building usable generative large language models (LLMs). While traditional RLHF-based alignment methods explicitly maximize the expected rewards from a separate reward model, more recent supervised alignment metho... | [
{
"id": "cvN9TGmzw1",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 4,
"summary": "The paper proposes an extension of Fisch et al 2024 work on Reward Distillation involving a regularization objective and demonstrate marginal improvements in term... | {
"rating": "3;3;5;5;6",
"rating_avg": 4.4,
"confidence": "4;4;3;4;4",
"confidence_avg": 3.8,
"soundness": "1;2;3;3;3",
"soundness_avg": 2.4,
"contribution": "2;2;2;3;2",
"contribution_avg": 2.2,
"presentation": "3;2;2;3;4",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.396733"
} | {
"id": "IdraKUcTb3",
"metareview": "This paper provides a new loss for simultaneous reward distillation and preference learning. The main paper is well-written and the experimental evidence is strong further motivating the proposal. However, the theoretical derivations are hard to follow, containing steps that are... | {
"decision": "Reject"
} |
l9Q9GtNwkT | 2405.16574v1 | Local Curvature Descent: Squeezing More Curvature out of Standard and Polyak Gradient Descent | {
"content": "## Abstract\n\nAbstract We contribute to the growing body of knowledge on more powerful and adaptive stepsizes for convex optimization, empowered by local curvature information . We do not go the route of fully-fledged second-order methods which require the expensive computation of the Hessian. Instead,... | [
{
"id": "gAr3qu42A0",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "- The paper introduces a new class of functions called convex and smooth with local structure. Many interesting examples are provided that satisfy the proposed as... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "5;4;4;3",
"confidence_avg": 4,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "1;1;2;3",
"contribution_avg": 1.75,
"presentation": "3;2;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.398136"
} | {
"id": "R2g4QWVPCd",
"metareview": "The paper studies settings of smooth convex optimization where additional information about \"local curvature\" of the objective is readily available. The paper proposes three variations of gradient descent -- generalizing GD with fixed step size and with Polyak step size, and a... | {
"decision": "Reject"
} |
lAXlDAdan5 | 2410.05911v1 | Accelerating Error Correction Code Transformers | {
"content": "## Abstract\n\nAbstract Error correction codes (ECC) are crucial for ensuring reliable information transmission in communication systems. Choukroun & Wolf ( 2022b ) recently introduced the Error Correction Code Transformer (ECCT), which has demonstrated promising performance across various transmission ... | [
{
"id": "3hUUFirpgc",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces simplifications to the Error Correction Code Transformer (ECCT) neural decoder model to reduce the memory and compute requirements. The auth... | {
"rating": "3;3;5;5;6;8",
"rating_avg": 5,
"confidence": "5;4;3;4;4;2",
"confidence_avg": 3.6666666666666665,
"soundness": "3;2;3;3;3;3",
"soundness_avg": 2.8333333333333335,
"contribution": "2;2;3;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;2;4;3;3;3",
"presentation_avg": 2.833333333333333... | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.399452"
} | {
"id": "ykgTRcBZoS",
"metareview": "This paper presents a method to enhance the efficiency and practicality of transformer-based error correction code decoders, focusing on the ECCT model introduced in prior work. Specifically, the authors propose: (i) an adaptive layer-based quantization scheme, (ii) masking atte... | {
"decision": "Reject"
} |
lBOvXyzQis | 2410.14556v1 | Measuring Diversity: Axioms and Challenges | {
"content": "## Abstract\n\nAbstract The concept of diversity is widely used in various applications: from image or molecule generation to recommender systems. Thus, being able to properly measure diversity is important. This paper addresses the problem of quantifying diversity for a set of objects. First, we make a... | [
{
"id": "v0vbfM4eSA",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper discusses some theoretical aspects of various diversity measures. It suggests that common diversity measures, such as Vendi Score and Determinantal Poin... | {
"rating": "5;5;5;5;5;8",
"rating_avg": 5.5,
"confidence": "3;2;3;3;3;3",
"confidence_avg": 2.8333333333333335,
"soundness": "2;3;3;2;3;4",
"soundness_avg": 2.8333333333333335,
"contribution": "2;2;2;2;2;3",
"contribution_avg": 2.1666666666666665,
"presentation": "2;3;3;3;3;2",
"presentation_avg": ... | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.400242"
} | {
"id": "Iss8eW1XYo",
"metareview": "The paper discusses some theoretical aspects of various diversity measures. Common diversity measures, e.g., Vendi Score and Determinantal Point Process scores, are optimised for computational simplicity rather than axiomatic optimal. They have potential drawbacks breaking some ... | {
"decision": "Reject"
} |
lBxJH2aTJr | 2410.02628v1 | Inverse Entropic Optimal Transport Solves Semi-supervised Learning via Data Likelihood Maximization | {
"content": "## Abstract\n\nAbstract Learning conditional distributions π ∗ ( ⋅ | x ) \\pi^{*}(\\cdot|x) italic_π start_POSTSUPERSCRIPT ∗ end_POSTSUPERSCRIPT ( ⋅ | italic_x ) is a central problem in machine learning, which is typically approached via supervised methods with paired data ( x , y ) ∼ π ∗ similar-to 𝑥 ... | [
{
"id": "dhCoB3s5Nx",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper introduces a novel semi-supervised learning approach that leverages both paired and unpaired data to learn conditional distributions through a data lik... | {
"rating": "3;5;6",
"rating_avg": 4.666666666666667,
"confidence": "3;4;4",
"confidence_avg": 3.6666666666666665,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "2;3;2",
"presentation_avg": 2.33333333333333... | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.401035"
} | {
"id": "ob20I5y1qJ",
"metareview": "This paper proposes a semi-supervised approach for domain translation that combines paired and unpaired data within a likelihood maximization framework. The authors establish a connection between their proposed loss function and inverse entropic optimal transport, introducing a ... | {
"decision": "Reject"
} |
lCasyP21Bf | 2404.18624v2 | Do Vision & Language Decoders use Images and Text equally? How Self-consistent are their Explanations? | {
"content": "## Abstract\n\nAbstract Vision and language model (VLM) decoders are currently the best-performing architectures on multimodal tasks. Next to predictions, they can also produce explanations, either in post-hoc or CoT settings.\nHowever, it is not clear how much they use the vision and text modalities wh... | [
{
"id": "2v88erufl5",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper investigates how VLMs balance text and image data when generating answers and explanations, evaluating whether their reliance on each modality shifts d... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "3;4;3;2",
"confidence_avg": 3,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;4;3;3",
"contribution_avg": 3,
"presentation": "3;4;3;2",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.401886"
} | {
"id": "8uMHx5tl7n",
"metareview": "### Summary:\nThis paper investigates how VLMs use text and image modalities when generating predictions and explanations, evaluating their self-consistency in both post-hoc and CoT settings. The key findings are that VLMs rely more heavily on text than images for predictions, b... | {
"decision": "Accept (Poster)"
} |
lE9s40eZgJ | 2406.08707v1 | mOSCAR: A Large-scale Multilingual and Multimodal Document-Level Corpus | {
"content": "## Abstract\n\nAbstract Multimodal Large Language Models (mLLMs) are trained on a large amount of text-image data. While most mLLMs are trained on caption-like data only, Alayrac et al. ( 2022 ) showed that additionally training them on interleaved sequences of text and images can lead to the emergence ... | [
{
"id": "fHVOZRZtW2",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper yields a powerful Multimodal Multilingual pretraining dataset for MLLM's. They choose a standard set of filters and approaches to arrive a higher quali... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "4;4;3;4",
"confidence_avg": 3.75,
"soundness": "3;3;3;4",
"soundness_avg": 3.25,
"contribution": "3;3;3;4",
"contribution_avg": 3.25,
"presentation": "3;4;3;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.402682"
} | {
"id": "HuGcYOtLNV",
"metareview": "This paper presents mOSCAR, described as the first large-scale multilingual and multimodal web-crawled document corpus. The key claims are that it covers 163 languages with 303M documents, 200B tokens and 1.15B images, and that models trained on mOSCAR show improved few-shot lea... | {
"decision": "Reject"
} |
lEsNGN1SjG | 2111.04404v2 | Information-theoretically Safe Bias Classifier Against Adversarial Attacks | {
"content": "## Abstract\n\nAbstract In this paper, the bias classifier is introduced, that is, the bias part\nof a DNN with Relu as the activation function is used as a classifier.\nThe work is motivated by the fact that\nthe bias part is a piecewise constant function with zero gradient and hence cannot be\ndirectl... | [
{
"id": "Xs9TKb3QCn",
"initial_rating": 1,
"confidence": 5,
"soundness": 1,
"contribution": 1,
"presentation": 1,
"summary": "This paper designs a \"provable\" adversarial robustness enhancing method named \"bias classifier\" that is supposed to help ML models achieve high adversarial ac... | {
"rating": "1;1;3;3",
"rating_avg": 2,
"confidence": "4;5;4;4",
"confidence_avg": 4.25,
"soundness": "1;1;2;2",
"soundness_avg": 1.5,
"contribution": "2;1;2;2",
"contribution_avg": 1.75,
"presentation": "2;1;1;2",
"presentation_avg": 1.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.403483"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
lFijzkTUNB | 2407.01976v2 | A Bounding Box is Worth One Token: Interleaving Layout and Text in a Large Language Model for Document Understanding | {
"content": "## Abstract\n\nAbstract Recently, many studies have demonstrated that exclusively incorporating OCR-derived text and spatial layouts with large language models (LLMs) can be highly effective for document understanding tasks. However, existing methods that integrate spatial layouts with text have limitat... | [
{
"id": "SuXNfTaSEx",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper presents LayTexLLM, which interleaves layout and text in an LLM for document understanding. There are two key ideas:\n- Spatial layout projector (SLP),... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "1;2;3;3",
"contribution_avg": 2.25,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.404243"
} | {
"id": "qLQ1Vq6jhS",
"metareview": "This paper introduces LayTextLLM, a method for incorporating document layout into LLMs. It uses: (1) a Spatial Layout Projector (SLP) to embed bounding box coordinates from OCR output, enabling layout-aware next token prediction; and (2) Shuffled-OCR supervised fine-tuning (SSFT... | {
"decision": "Reject"
} |
lGDmwb12Qq | 2410.06527v1 | The Sampling-Gaussian for stereo matching | {
"content": "## Abstract\n\nAbstract The soft-argmax operation is widely adopted in neural network-based stereo matching methods to enable differentiable regression of disparity. However, network trained with soft-argmax is prone to being multimodal due to absence of explicit constraint to the shape of the probabili... | [
{
"id": "y7Per0nrJU",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 1,
"presentation": 1,
"summary": "This paper proposed ‘Sampling Gaussian’ as supervision for disparity estimation. Moreover, this paper also studied the issues of using soft-argmax for depth estim... | {
"rating": "3;5;6",
"rating_avg": 4.666666666666667,
"confidence": "4;4;5",
"confidence_avg": 4.333333333333333,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "1;2;3",
"contribution_avg": 2,
"presentation": "1;3;3",
"presentation_avg": 2.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.405065"
} | {
"id": "lN6YMauuU3",
"metareview": "The paper introduces an approach called Sampling-Gaussian to addresses the shortcomings of the soft-argmax operation often encountered in stereo matching. The method can be applied universally to soft-argmax-based methods to better tackle multimodal distributions. The major conc... | {
"decision": "Reject"
} |
lGWaAIC9gU | 2409.15199v1 | Learning from Contrastive Prompts: Automated Optimization and Adaptation | {
"content": "## Abstract\n\nAbstract As LLMs evolve, significant effort is spent on manually crafting prompts. While existing prompt optimization methods automate this process, they rely solely on learning from incorrect samples, leading to a sub-optimal performance. Additionally, an unexplored challenge in the lite... | [
{
"id": "jdHitAJ3Si",
"initial_rating": 5,
"confidence": 2,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The authors propose an engineering technique for prompt tuning that leverages an LLM's reasoning and generation capabilities. To tune prompts using the LLM, they ... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;4;2;3",
"confidence_avg": 3.25,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "1;2;3;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.405686"
} | {
"id": "5P8PFq0Nah",
"metareview": "The paper proposes a new prompt optimization method which is inspired by contrastive learning. The method makes use of both correct and incorrect samples in prompt optimization. The paper also proposes a method for achieving prompt adaptation across different models and language... | {
"decision": "Reject"
} |
lHSeDYamnz | 2410.16454v1 | Catastrophic Failure of LLM Unlearning via Quantization | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) have shown remarkable proficiency in generating text, benefiting from extensive training on vast textual corpora. However, LLMs may also acquire unwanted behaviors from the diverse and sensitive nature of their training data, which can include copyrig... | [
{
"id": "FaaCo636JK",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 2,
"summary": "This paper addresses the problems that arise when quantizing models after unlearning. The authors also propose a quatization-robust unlearning algorithm.",
"s... | {
"rating": "3;3;5;6;6",
"rating_avg": 4.6,
"confidence": "4;4;4;3;4",
"confidence_avg": 3.8,
"soundness": "2;2;3;3;3",
"soundness_avg": 2.6,
"contribution": "2;3;2;3;2",
"contribution_avg": 2.4,
"presentation": "3;2;2;1;3",
"presentation_avg": 2.2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.406413"
} | {
"id": "JDHu2Arvfb",
"metareview": "This paper examines LLMs after unlearning, a process that removes certain knowledge from the model. The authors identify a critical failure mode: when the weights of an unlearned LLM are quantized, the supposedly removed knowledge can be recovered. They then propose a saliency-b... | {
"decision": "Accept (Poster)"
} |
lIVRgt4nLv | 2410.08164v1 | Agent S: An Open Agentic Framework that Uses Computers Like a Human | {
"content": "## Abstract\n\nAbstract We present Agent S, an open agentic framework that enables autonomous interaction with computers through a Graphical User Interface (GUI), aimed at transforming human-computer interaction by automating complex, multi-step tasks. Agent S aims to address three key challenges in aut... | [
{
"id": "q51Qr6zHIZ",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces Agent S, a groundbreaking framework designed to automate complex, multi-step tasks on computers through Graphical User Interface (GUI) intera... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "3;4;3;5",
"confidence_avg": 3.75,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;1;3;3",
"contribution_avg": 2.25,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.407208"
} | {
"id": "QVaFf1qoI6",
"metareview": "This paper introduces Agent S, a novel framework for controlling GUI-based operating systems, leveraging three core strategies: experience-augmented hierarchical planning, continual memory update, and an Agent-Computer Interface. \n\nThe proposed framework is comprehensive and i... | {
"decision": "Accept (Poster)"
} |
lJ66m0ibQL | 2410.06084v1 | Diversity-Rewarded CFG Distillation | {
"content": "## Abstract\n\nAbstract Generative models are transforming creative domains such as music generation, with inference-time strategies like Classifier-Free Guidance (CFG) playing a crucial role.\nHowever, CFG doubles inference cost while limiting originality and diversity across generated contents.\nIn th... | [
{
"id": "2eVdowZJtj",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces a fine-tuning method that enhances both quality and diversity in generative models, targeting CFG’s limitations in computational cost and di... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "4;3;3;4",
"confidence_avg": 3.5,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;3;3;2",
"contribution_avg": 2.5,
"presentation": "3;3;4;4",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.407944"
} | {
"id": "vFrGXloALm",
"metareview": "This paper introduces a fine-tuning method that enhances both quality and diversity in generative models, targeting classifier-free guidance’s limitations in computational cost and diversity reduction. \n\nAll reviewers described originality, clear description, and promising res... | {
"decision": "Accept (Poster)"
} |
lJGEyRSXDJ | 2410.22071v1 | Distinguishing Ignorance from Error in LLM Hallucinations | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) are susceptible to hallucinations—outputs that are ungrounded, factually incorrect, or inconsistent with prior generations. We focus on close-book Question Answering (CBQA), where previous work has not fully addressed the distinction between two possi... | [
{
"id": "mvndIjdIPH",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes a method to separate two causes for LLM hallucination – when the model does not have the related parametric knowledge, and when the model lies ... | {
"rating": "3;3;5;5;6",
"rating_avg": 4.4,
"confidence": "4;4;5;3;4",
"confidence_avg": 4,
"soundness": "2;2;3;3;3",
"soundness_avg": 2.6,
"contribution": "2;2;2;2;3",
"contribution_avg": 2.2,
"presentation": "1;2;3;3;3",
"presentation_avg": 2.4
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.408632"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
lJWHOT1ZAV | 2402.05558v1 | Flashback: Understanding and Mitigating Forgetting in Federated Learning | {
"content": "## Abstract\n\nAbstract In Federated Learning (FL), forgetting, or the loss of knowledge across rounds, hampers algorithm convergence, particularly in the presence of severe data heterogeneity among clients.\nThis study explores the nuances of this issue, emphasizing the critical role of forgetting in F... | [
{
"id": "QfWPjbTd7k",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper formulates the NonIID problem in FL as the catastrophic forgetting from both local and global perspective. To solve this problem, they propose using kn... | {
"rating": "3;3;3;5;5",
"rating_avg": 3.8,
"confidence": "4;5;4;4;5",
"confidence_avg": 4.4,
"soundness": "2;2;2;3;2",
"soundness_avg": 2.2,
"contribution": "2;1;2;2;2",
"contribution_avg": 1.8,
"presentation": "2;2;3;3;2",
"presentation_avg": 2.4
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.409298"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
lJcSDsGgYH | 2406.04370v2 | Large Language Model Confidence Estimation via Black-Box Access | {
"content": "## Abstract\n\nAbstract Estimating uncertainty or confidence in the responses of a model can be significant in evaluating trust not only in the responses, but also in the model as a whole. In this paper, we explore the problem of estimating confidence for responses of large language models (LLMs) with s... | [
{
"id": "4t0rohuX0Z",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper studies the problem of uncertainty quantification in language model responses using only black-box access to the language model. In order to quantify u... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;3;3;3",
"confidence_avg": 3.25,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;2;2",
"contribution_avg": 2.25,
"presentation": "2;3;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.409947"
} | {
"id": "EuNA4sV3mF",
"metareview": "The authors propose a method for calibrating LLM outputs given black-box query access. Their method involves 1) obtaining a set of generations by introducting heuristics perturbations in the input prompt (e.g., through paraphrasing) as well as directly in the output, 2) extracti... | {
"decision": "Reject"
} |
lJdgUUcLaA | 2410.02666v1 | AlphaIntegrator: Transformer Action Search for Symbolic Integration Proofs | {
"content": "## Abstract\n\nAbstract We present the first correct-by-construction learning-based system for step-by-step mathematical integration. The key idea is to learn a policy, represented by a GPT transformer model, which guides the search for the right mathematical integration rule, to be carried out by a sym... | [
{
"id": "XHHmqu7Hze",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 4,
"summary": "The paper presents AlphaIntegrator -- a learning-based system for computing antiderivatives of mathematical expressions in a step-by-step manner. The system is a ... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "3;3;3;4",
"confidence_avg": 3.25,
"soundness": "1;2;2;3",
"soundness_avg": 2,
"contribution": "1;3;2;2",
"contribution_avg": 2,
"presentation": "2;2;2;4",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.410743"
} | {
"id": "NhRuBO2x8Y",
"metareview": "This paper presents, AlphaIntegrator, which is a decoder-only transformer model trained to predict stepwise integration rules for computing antiderivatives of mathematical expressions. Groundtruth integration rules are collected by running SymPy over a synthetic dataset collecte... | {
"decision": "Reject"
} |
lJpqxFgWCM | 2410.03825v1 | MonST3R: A Simple Approach for Estimating Geometry in the Presence of Motion | {
"content": "## Abstract\n\nAbstract Estimating geometry from dynamic scenes, where objects move and deform over time, remains a core challenge in computer vision. Current approaches often rely on multi-stage pipelines or global optimizations that decompose the problem into subtasks, like depth and flow, leading to ... | [
{
"id": "m72rackV87",
"initial_rating": 8,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper first identifies the limitations in joint pose and geometry estimation methods on dynamic scenes, and proposes to finetune based on the recent geometric... | {
"rating": "5;5;8;8;8",
"rating_avg": 6.8,
"confidence": "5;4;4;5;5",
"confidence_avg": 4.6,
"soundness": "3;3;4;4;3",
"soundness_avg": 3.4,
"contribution": "3;2;3;4;3",
"contribution_avg": 3,
"presentation": "3;2;3;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.411564"
} | {
"id": "BWvU4ihPUG",
"metareview": "The paper present an approach to depth and pose estimation from monocular video.\nIt extends DUST3R to handle dynamic scenes by fine-tuning on appropriate datasets and tailoring its global pose optimization to dynamic scenes. Specifically, a sliding-window based optimization imp... | {
"decision": "Accept (Spotlight)"
} |
lNtio1tdbL | 2411.03055v2 | ATM: Improving Model Merging by Alternating Tuning and Merging | {
"content": "## Abstract\n\nAbstract Model merging has recently emerged as a cost-efficient paradigm for multi-task learning. Among current approaches, task arithmetic (Ilharco et al., 2022 ) stands out for its simplicity and effectiveness.\nIn this paper, we motivate the effectiveness of task vectors by linking the... | [
{
"id": "68DdDgwuln",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposes a novel model “merging” algorithm, where the goal is to merge separately tuned, task-specific models into a joint multi-task model. The paper ... | {
"rating": "1;1;5;5",
"rating_avg": 3,
"confidence": "5;5;4;4",
"confidence_avg": 4.5,
"soundness": "1;1;3;3",
"soundness_avg": 2,
"contribution": "1;1;3;2",
"contribution_avg": 1.75,
"presentation": "1;1;3;2",
"presentation_avg": 1.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.412366"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
lOi6FtIwR8 | 2405.13967v4 | Model Editing as a Robust and Denoised variant of DPO: A Case Study on Toxicity | {
"content": "## Abstract\n\nAbstract Recent alignment algorithms such as direct preference optimization (DPO) have been developed to improve the safety of large language models (LLMs) by training these models to match human behaviors exemplified by preference data.\nHowever, these methods are\nboth computationally i... | [
{
"id": "BTb0CpwLN2",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper presents a novel flow matching method. This method can be applied to high dimensional data measured at non-equidistant time points without reducing dime... | {
"rating": "6;6",
"rating_avg": 6,
"confidence": "4;1",
"confidence_avg": 2.5,
"soundness": "4;3",
"soundness_avg": 3.5,
"contribution": "4;2",
"contribution_avg": 3,
"presentation": "4;3",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.413351"
} | {
"id": "d8DalIKkNQ",
"metareview": "This paper provides a tuning-free method called ProFS as an alternative to direct preference optimization (DPO) to reduce model toxicity. Methods like DPO aim to make large language models (LLMs) safer by training them to mimic human preferences. However, these methods are comp... | {
"decision": "Accept (Poster)"
} |
lPJUQsSIxm | 2408.15231v1 | DCT-CryptoNets: Scaling Private Inference in the Frequency Domain | {
"content": "## Abstract\n\nAbstract The convergence of fully homomorphic encryption (FHE) and machine learning offers unprecedented opportunities for private inference of sensitive data. FHE enables computation directly on encrypted data, safeguarding the entire machine learning pipeline, including data and model c... | [
{
"id": "KWS07xwI8B",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces DCT-CryptoNets, which performs neural network (CNN) inference entirely using TFHE-based homomorphic encryption in the frequency domain. Unli... | {
"rating": "5;6;8",
"rating_avg": 6.333333333333333,
"confidence": "4;3;4",
"confidence_avg": 3.6666666666666665,
"soundness": "3;3;4",
"soundness_avg": 3.3333333333333335,
"contribution": "2;3;4",
"contribution_avg": 3,
"presentation": "3;3;4",
"presentation_avg": 3.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.414357"
} | {
"id": "8Ahy3nHGFT",
"metareview": "The paper proposes a new technique DCT-Cryptonets in order to be able to perform cryptographically private (using Fully Homomorphic Encryption) inference on neural networks, particularly for convolutional neural networks. The main contribution is the method to perform operations... | {
"decision": "Accept (Poster)"
} |
lQYi2zeDyh | 2405.16924v1 | Demystifying amortized causal discovery with transformers | {
"content": "## Abstract\n\nAbstract Supervised learning approaches for causal discovery from observational data often achieve competitive performance despite seemingly avoiding explicit assumptions that traditional methods make for identifiability. In this work, we investigate CSIvA [ 1 ] , a transformer-based mode... | [
{
"id": "6ilmVqPacX",
"initial_rating": 6,
"confidence": 3,
"soundness": 4,
"contribution": 3,
"presentation": 3,
"summary": "This paper analyzes amortized causal discovery, where synthetic data allows supervised learning of causal discovery. The investigation focuses on whether learning... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "2;2;3;4",
"soundness_avg": 2.75,
"contribution": "1;2;3;3",
"contribution_avg": 2.25,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.415239"
} | {
"id": "aHJl9bvyZ3",
"metareview": "While the manuscript targets an important question that why amortized causal discovery with transformers works, major concerns raised by the reviewers—particularly regarding the limited scope of experiments (mostly bivariate) and the lack of theoretical claims—still remain. \n\n... | {
"decision": "Reject"
} |
lR7rqLtsXZ | 2410.01623v2 | Fira: Can We Achieve Full-rank Training of LLMs Under Low-rank Constraint? | {
"content": "## Abstract\n\nAbstract Low-rank training has emerged as a promising approach for reducing memory usage in training Large Language Models (LLMs). Previous methods either rely on decomposing weight matrices (e.g., LoRA), or seek to decompose gradient matrices (e.g., GaLore) to ensure reduced memory consu... | [
{
"id": "acVpS1PsvH",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The paper proposes a memory-efficient training framework called Fira for the pre-training and fine-tuning of large language models (LLMs). Fira aims to reconcile ... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "4;3;4;4",
"confidence_avg": 3.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;2;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.415928"
} | {
"id": "wonwJn5gUp",
"metareview": "This paper proposes a new strategy for training LLMs, where they use the full rank update but maintain optimizer stats in a low-rank fashion to save memory. Two main techniques: Norm-Based Scaling and the Norm-Growth Limiter are proposed to form the actual updates using low-rank... | {
"decision": "Reject"
} |
lS2SGfWizd | 2410.14919v2 | Adversarial Score Identity Distillation: Rapidly Surpassing the Teacher in One Step | {
"content": "## Abstract\n\nAbstract Score identity Distillation (SiD) is a data-free method that has achieved state-of-the-art performance in image generation by leveraging only a pretrained diffusion model, without requiring any training data. However, the ultimate performance of SiD is constrained by the accuracy... | [
{
"id": "YyNOpbqyHp",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "The authors propose SiDA (Score identity Distillation with Adversarial Loss), an enhanced version of Score identity Distillation (SiD) that aims to improve diffus... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "4;4;4;3",
"confidence_avg": 3.75,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "2;3;2;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.416693"
} | {
"id": "K7z93FEAOQ",
"metareview": "This paper improves score identity distillation by integrating real data with adversarial loss, which enhances generation quality even surpassing the teacher in one-step generation, while also improving distillation efficiency. Overall, this work receives positive scores, and th... | {
"decision": "Accept (Poster)"
} |
lTrrnNdkOX | 2409.04005v2 | Qihoo-T2X: An Efficient Proxy-Tokenized Diffusion Transformer for Text-to-Any-Task | {
"content": "## Abstract\n\nAbstract The global self-attention mechanism in diffusion transformers involves redundant computation due to the sparse and redundant nature of visual information, and the attention map of tokens within a spatial window shows significant similarity.\nTo address this redundancy, we propose... | [
{
"id": "w0JpU2AZfe",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper proposes Proxy-Tokenized Diffusion Transformer to reduce the computational complexity. By utilizing sparse representative token attention, window and s... | {
"rating": "5;6;6;6;8",
"rating_avg": 6.2,
"confidence": "4;5;3;3;5",
"confidence_avg": 4,
"soundness": "2;3;3;3;3",
"soundness_avg": 2.8,
"contribution": "2;3;3;3;4",
"contribution_avg": 3,
"presentation": "3;3;2;3;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.417419"
} | {
"id": "uh86twssfU",
"metareview": "(a) The paper proposes the Proxy-Tokenized Diffusion Transformer to reduce computational complexity by using sparse representative token attention and window-based attention. The Qihoo-T2X family includes models for text-to-image, text-to-video, and text-to-multi-view tasks.\n\n... | {
"decision": "Accept (Poster)"
} |
lVp97zZ5i8 | 2409.01156v1 | TempMe: Video Temporal Token Merging for Efficient Text-Video Retrieval | {
"content": "## Abstract\n\nAbstract Most text-video retrieval methods utilize the text-image pre-trained CLIP as a backbone, incorporating complex modules that result in high computational overhead. As a result, many studies focus on efficient fine-tuning. The primary challenge in efficient adaption arises from the... | [
{
"id": "HUogMAe7zJ",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "To enhance the efficiency of text-video retrieval, this paper introduces Temporal Token Merging (TempMe), a parameter- and inference-efficient architecture aimed ... | {
"rating": "5;6;6",
"rating_avg": 5.666666666666667,
"confidence": "5;3;3",
"confidence_avg": 3.6666666666666665,
"soundness": "3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3",
"contribution_avg": 2.6666666666666665,
"presentation": "2;2;4",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.418145"
} | {
"id": "wCE4nIIF3d",
"metareview": "This paper proposed a new temporal token merging method for text-video retrieval to educe temporal redundancy and training/inference cost. The motivation is reasonable and the experiments are comprehensive, demonstrating efficiency, effectiveness, and generalization capabilities... | {
"decision": "Accept (Poster)"
} |
lWGXftRS5h | 2410.21273v1 | On Inductive Biases That Enable Generalization in Diffusion Transformers | {
"content": "## Abstract\n\nAbstract Recent work studying the generalization of diffusion models with UNet-based denoisers reveals\ninductive biases that can be expressed via geometry-adaptive harmonic bases.\nHowever, in practice, more recent denoising networks are often based on transformers, e . g ., the diffusio... | [
{
"id": "A6dFL4C30h",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper investigates the inductive bias of diffusion transformers. Specifically, the authors start by looking at the difference in the generalization behavior ... | {
"rating": "3;3;5;8",
"rating_avg": 4.75,
"confidence": "3;3;3;2",
"confidence_avg": 2.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;3;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.418769"
} | {
"id": "CJCsRmJVvM",
"metareview": "This paper explores the generalization behaviour of Diffusion Transformers, and claims that attention locality is the inductive bias, as opposed to UNet-based diffusion models which were said to have harmonic bases as inductive bias in prior work. Local attention windows are pro... | {
"decision": "Reject"
} |
lXv9DTw650 | 2409.17564v1 | General Compression Framework for Efficient Transformer Object Tracking | {
"content": "## Abstract\n\nAbstract Transformer-based trackers have established a dominant role in the field of visual object tracking. While these trackers exhibit promising performance, their deployment on resource-constrained devices remains challenging due to inefficiencies. To improve the inference efficiency ... | [
{
"id": "PRvaWLWkZP",
"initial_rating": 5,
"confidence": 5,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "In this paper, the authors proposed a general model compression framework for efficient Transformer object tracking, named CompressTracker. The method adopts a no... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "3;4;5;3",
"confidence_avg": 3.75,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "3;3;2;3",
"contribution_avg": 2.75,
"presentation": "2;3;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.419697"
} | {
"id": "tRVTiGnP4i",
"metareview": "This paper presents a model compression framework aimed at enhancing the efficiency of transformer-based object tracking. It received mixed reviews, with scores of 5, 5, 6, and 6, resulting in an average score of 5.5.\n\nReviewers acknowledged the paper’s strengths, particularly... | {
"decision": "Reject"
} |
lYDiuQ7vJA | 2405.16606v1 | Link Prediction on Textual Edge Graphs | {
"content": "## Abstract\n\nAbstract Textual-edge Graphs (TEGs), characterized by rich text annotations on edges, are increasingly significant in network science due to their ability to capture rich contextual information among entities. Existing works have proposed various edge-aware graph neural networks (GNNs) or... | [
{
"id": "ngA9L520CK",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper studies link prediction on textual-edge graphs (TEGs). They proposed a framework LINK2DOC, which models TEGs from two perspectives, i.e., the text view... | {
"rating": "3;5;5;5;5",
"rating_avg": 4.6,
"confidence": "3;3;4;4;4",
"confidence_avg": 3.6,
"soundness": "1;3;3;3;2",
"soundness_avg": 2.4,
"contribution": "1;3;3;2;2",
"contribution_avg": 2.2,
"presentation": "1;3;3;2;2",
"presentation_avg": 2.2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.420289"
} | {
"id": "Sa6H9J2qbu",
"metareview": "Reviewers find that the problem of LP on textual edge graphs is well motivated with solid analysis and example in Figure 1. The transition graph idea is interesting.\n\nHowever, experimental setup requires further clarification, and the presentation and organization of the paper... | {
"decision": "Reject"
} |
lYRkGZZi9D | 2402.04051v4 | Analysis of Linear Mode Connectivity via Permutation-Based Weight Matching: With Insights into Other Permutation Search Methods | {
"content": "## Abstract\n\nAbstract Recently, Ainsworth et al. ( 2023 ) showed that using weight matching (WM) to minimize the L 2 subscript 𝐿 2 L_{2} italic_L start_POSTSUBSCRIPT 2 end_POSTSUBSCRIPT distance in a permutation search of model parameters effectively identifies permutations that satisfy linear mode c... | [
{
"id": "87jIyqBA0L",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper analyses the the weight matching algorithm through singular value decomposition. \n\nPaper starts with the observation that algorithms like weight matc... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "3;4;4;3",
"confidence_avg": 3.5,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;3;4;4",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.421453"
} | {
"id": "tSCrmKvR9A",
"metareview": "The paper 'Analysis of Linear Mode Connectivity via Permutation-Based Weight Matching: With Insights into Other Permutation Search Methods' was reviewed by 4 reviewers who gave it an average score of 7.0 (final scores: 6+6+8+8). The reviewers found this work relevant for the con... | {
"decision": "Accept (Poster)"
} |
lYXhiCYkPn | 2410.10241v1 | Revisiting and Benchmarking Graph Autoencoders: A Contrastive Learning Perspective | {
"content": "## Abstract\n\nAbstract Graph autoencoders (GAEs) are self-supervised learning models that can learn meaningful representations of graph-structured data by reconstructing the input graph from a low-dimensional latent space. Over the past few years, GAEs have gained significant attention in academia and ... | [
{
"id": "VjvuwbMcSW",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper analyzes the connection between two graph self-supervised learning paradigms, i.e., graph contrastive learning and graph autoencoders. Then, it enhances... | {
"rating": "3;3;5;5;6",
"rating_avg": 4.4,
"confidence": "4;4;4;4;5",
"confidence_avg": 4.2,
"soundness": "2;3;3;3;3",
"soundness_avg": 2.8,
"contribution": "2;2;2;1;3",
"contribution_avg": 2,
"presentation": "3;3;3;2;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.422582"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
lYongcxaNz | 2410.21698v1 | On the Role of Depth and Looping for In-Context Learning with Task Diversity | {
"content": "## Abstract\n\nAbstract The intriguing in-context learning (ICL) abilities of deep Transformer models have lately garnered significant attention. By studying in-context linear regression on unimodal Gaussian data, recent empirical and theoretical works have argued that ICL emerges from Transformers’ abi... | [
{
"id": "Hzy5WHWdQQ",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper studies in-context learning (ICL) for multiple regression tasks, focusing on \"task diversity\" by varying the covariances of linear regression tasks. ... | {
"rating": "3;3;5;6;8",
"rating_avg": 5,
"confidence": "4;3;3;4;3",
"confidence_avg": 3.4,
"soundness": "3;2;3;3;4",
"soundness_avg": 3,
"contribution": "1;2;2;3;3",
"contribution_avg": 2.2,
"presentation": "2;1;2;3;3",
"presentation_avg": 2.2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.423518"
} | {
"id": "jVi0poac9u",
"metareview": "This paper considers learning ability of Transformers to perform in-context learning where the covariance of the input can differ in each task. It is shown that Transformers can implement multi-step gradient descent. In addition to that, the authors discuss robustness of Transfo... | {
"decision": "Reject"
} |
lbe3BiDCQr | 2406.02524v2 | CheckEmbed: Effective Verification of LLM Solutions to Open-Ended Tasks | {
"content": "## Abstract\n\nAbstract Large Language Models (LLMs) are revolutionizing various domains, yet verifying their answers remains a significant challenge, especially for intricate open-ended tasks such as consolidation, summarization, and extraction of knowledge.\nIn this work, we propose CheckEmbed : an ac... | [
{
"id": "1LSTUkOsmy",
"initial_rating": 3,
"confidence": 4,
"soundness": 1,
"contribution": 1,
"presentation": 2,
"summary": "This paper addresses LLM hallucination detection and answer verification. Following the assumption of SelfCheckGPT, the authors propose sampling multiple response... | {
"rating": "3;3;3;3",
"rating_avg": 3,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "3;2;3;1",
"soundness_avg": 2.25,
"contribution": "3;2;2;1",
"contribution_avg": 2,
"presentation": "2;3;3;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.424372"
} | {
"id": "D6MrJm7tUf",
"metareview": "**Summary:** \n\nThis work highlights the limitations of SelfCheckGPT and BERTScore, which are models designed to evaluate the similarity between two passages. To address these shortcomings, it introduces CheckEmbed, a novel approach that evaluates the embeddings of entire LLM-g... | {
"decision": "Reject"
} |
lbj0i29Z92 | 2407.19594v2 | Meta-Rewarding Language Models: Self-Improving Alignment with LLM-as-a-Meta-Judge | {
"content": "## Abstract\n\nAbstract Large Language Models (LLMs) are rapidly surpassing human knowledge in many domains. While improving these models traditionally relies on costly human data, recent self-rewarding mechanisms (Yuan et al., 2024c ) have shown that LLMs can improve by judging their own responses inst... | [
{
"id": "6yGjv5pesk",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This study presents an innovative Meta-Rewarding mechanism to enhance the existing self-rewarding framework. This new approach enables the model to evaluate its j... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.425029"
} | {
"id": "KdUDgwtOAP",
"metareview": "The paper proposes self improving with a meta judge.\n\nStrengths:\nTimely topic \nClear writing \nInteresting approach and techniques, with benefits (e.g. improving generation and verification at once)\n\nWeaknesses: \nThere were concerns regarding the compute necessary, and... | {
"decision": "Reject"
} |
ldVkAO09Km | 2405.20555v1 | Diffusion Actor-Critic: Formulating Constrained Policy Iteration as Diffusion Noise Regression for Offline Reinforcement Learning | {
"content": "## Abstract\n\nAbstract In offline reinforcement learning (RL), it is necessary to manage out-of-distribution actions to prevent overestimation of value functions. Policy-regularized methods address this problem by constraining the target policy to stay close to the behavior policy. Although several app... | [
{
"id": "llKqtEUeHY",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper integrates the advantage-weighted regression and diffusion model for offline reinforcement learning. This paper introduces the Diffusion Actor-Critic (... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "3;4;4;3",
"confidence_avg": 3.5,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.425678"
} | {
"id": "hfRjGSmaDS",
"metareview": "Diffusion Actor-Critic: Formulating Constrained Policy Iteration as Diffusion Noise Regression for Offline Reinforcement Learning \n\nSummary: The paper introduces Diffusion Actor-Critic (DAC), a new offline reinforcement learning algorithm that utilizes diffusion models to gene... | {
"decision": "Accept (Poster)"
} |
leACdxBEgv | 2405.16195v2 | Adaptive $Q$-Network: On-the-fly Target Selection for Deep Reinforcement Learning | {
"content": "## Abstract\n\nAbstract Deep Reinforcement Learning (RL) is well known for being highly sensitive to hyperparameters, requiring practitioners substantial efforts to optimize them for the problem at hand. This also limits the applicability of RL in real-world scenarios. In recent years, the field of auto... | [
{
"id": "Ack16Ja4mC",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The authors have developed a new AutoRL approach called Adaptive Q-Network (AdaQN), which aims to address the non-stationary optimization procedure in RL without ... | {
"rating": "5;6;8",
"rating_avg": 6.333333333333333,
"confidence": "3;3;3",
"confidence_avg": 3,
"soundness": "3;3;3",
"soundness_avg": 3,
"contribution": "3;3;3",
"contribution_avg": 3,
"presentation": "3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.426443"
} | {
"id": "DgIWClPYVY",
"metareview": "The paper presents an autoRL technique to dynamically and automatically select the best hyperparameters in temporal different Q-network updates. The strength of this work is an effective technique that clearly improves the stability and sample efficiency of learning. The main ... | {
"decision": "Accept (Poster)"
} |
lessla98Wp | 2410.04972v2 | L-C4: Language-Based Video Colorization for Creative and Consistent Colors | {
"content": "## Abstract\n\nAbstract Automatic video colorization is inherently an ill-posed problem because each monochrome frame has multiple optional color candidates.\nPrevious exemplar-based video colorization methods restrict the user’s imagination due to the elaborate retrieval process. Alternatively, conditi... | [
{
"id": "dt6GXheLSd",
"initial_rating": 5,
"confidence": 5,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposes a language-based video colorization model and employs temporally deformable attention and fusion module to handle inter-frame color consistenc... | {
"rating": "5;5;6;6;6",
"rating_avg": 5.6,
"confidence": "5;5;4;4;4",
"confidence_avg": 4.4,
"soundness": "2;3;3;3;3",
"soundness_avg": 2.8,
"contribution": "2;2;2;3;3",
"contribution_avg": 2.4,
"presentation": "2;3;3;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.427289"
} | {
"id": "wmmMZJcUad",
"metareview": "The submission introduces L-C4, a language-based video colorization method. The reviewers appreciated the novel integration of components addressing temporal consistency and creative flexibility. However, concerns were raised regarding computational efficiency, limited compariso... | {
"decision": "Reject"
} |
lfPkGWXLLf | 2409.19839v2 | ForecastBench: A Dynamic Benchmark of AI Forecasting Capabilities | {
"content": "## Abstract\n\nAbstract Forecasts of future events are essential inputs into informed decision-making. Machine learning (ML) systems have the potential to deliver forecasts at scale, but there is no framework for evaluating the accuracy of ML systems on a standardized set of forecasting questions. To ad... | [
{
"id": "ejXjEHgPm2",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "A dynamic dataset of forecasting qs about the future is aggregated and proposed as a benchmark for LLMs. Performance of a variety of LLMs is compared with human."... | {
"rating": "5;5;5;8;8;8",
"rating_avg": 6.5,
"confidence": "4;4;4;4;4;4",
"confidence_avg": 4,
"soundness": "2;3;2;3;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "3;3;2;3;4;3",
"contribution_avg": 3,
"presentation": "3;1;3;3;4;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.428098"
} | {
"id": "JZVayRLAHa",
"metareview": "This well-written and well-motivated paper proposes a new benchmark for LLM-based forecasting that aims to mitigate the staleness of existing benchmarks. It has been evaluated by 6 knowledgeable reviewers. Most of them recommended its acceptance (including 2 straight acceptances... | {
"decision": "Accept (Poster)"
} |
lgnAEBE1Xq | 2401.10458v1 | Contrastive Unlearning: A Contrastive Approach to Machine Unlearning | {
"content": "## Abstract\n\nAbstract Machine unlearning aims to eliminate the influence of a subset of training samples (i.e., unlearning samples) from a trained model. Effectively and efficiently removing the unlearning samples without negatively impacting the overall model performance is still challenging. In this... | [
{
"id": "mOI51cIRHS",
"initial_rating": 5,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "In this paper, the authors propose a new machine unlearning method, inspired by contrastive learning. \n\nThe proposed contrastive unlearning pushes away the posi... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "1;2;2;2",
"confidence_avg": 1.75,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "3;2;3;3",
"contribution_avg": 2.75,
"presentation": "2;1;2;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.428814"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
lgsyLSsDRe | 2405.17428v1 | NV-Embed: Improved Techniques for Training LLMs as Generalist Embedding Models | {
"content": "## Abstract\n\nAbstract Decoder-only large language model (LLM)-based embedding models are beginning to outperform BERT or T5-based embedding models in general-purpose text embedding tasks, including dense vector-based retrieval.\nIn this work, we introduce the NV-Embed model with a variety of architect... | [
{
"id": "THdbu5zx1W",
"initial_rating": 8,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper gives a summary of the NV-Embed model that achieved the top performance in the MTEB benchmark. \nThe techniques used are\n1. latent attention layer tha... | {
"rating": "6;6;8;8",
"rating_avg": 7,
"confidence": "4;3;3;5",
"confidence_avg": 3.75,
"soundness": "2;3;3;3",
"soundness_avg": 2.75,
"contribution": "3;2;3;3",
"contribution_avg": 2.75,
"presentation": "3;2;3;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.429383"
} | {
"id": "n2u15UGmMM",
"metareview": "The paper presents a suite of techniques to improve the general purpose embeddings produced by large decoder-only LLMs. This includes a latent attention layer to create the summary embedding, a contrastive training objective for instruction tuning, and careful creation of the tr... | {
"decision": "Accept (Spotlight)"
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.