paper_id string | arxiv_id string | title string | markdown dict | reviews list | scores dict | metadata dict | meta_review dict | decision dict |
|---|---|---|---|---|---|---|---|---|
lh0iTFCD1y | 2406.09864v2 | LUMA: A Benchmark Dataset for Learning from Uncertain and Multimodal Data | {
"content": "## Abstract\n\nAbstract Multimodal Deep Learning enhances decision-making by integrating diverse information sources, such as texts, images, audio, and videos. To develop trustworthy multimodal approaches, it is essential to understand how uncertainty impacts these models. We propose LUMA, a unique benc... | [
{
"id": "DeOXQE7c26",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper uses CIFAR image data, generates text and spoken word labels for these using a well described process, and creates a multimodal data set. It then descr... | {
"rating": "3;3;3;3",
"rating_avg": 3,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.429981"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
lidVssyB7G | 2406.19388v2 | Taming Data and Transformers for Audio Generation | {
"content": "## Abstract\n\nAbstract {NoHyper} * * footnotetext: Work partially done during an internship at Snap Inc. Generating ambient sounds is a challenging task due to data scarcity and often insufficient caption quality, making it difficult to employ large-scale generative models for the task. In this work, w... | [
{
"id": "m1kEgXDNKG",
"initial_rating": 3,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposes an audio captioning method called AutoCap, an audio generation model called GenAu and an audio dataset called AutoReCap-XL.",
"strengths":... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "5;5;5;4",
"confidence_avg": 4.75,
"soundness": "2;2;4;3",
"soundness_avg": 2.75,
"contribution": "2;2;3;2",
"contribution_avg": 2.25,
"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": "2026-01-14T22:16:03.431029"
} | {
"id": "eqPWbDtLEY",
"metareview": "> This paper introduces a high-quality and efficient audio captioning model, named AutoCap, which demonstrates improvements in generation quality while being four times faster than current state-of-the-art (SoTA) models. The authors further present AutoReCap-XL, a large-scale au... | {
"decision": "Reject"
} |
ljAS7cPAU0 | 2411.03753v1 | Symbolic regression via MDLformer-guided search: from minimizing prediction error to minimizing description length | {
"content": "## \\thesubsection Symbolic Regression on Black-Box Problems\n\n[24]r7cm\n\\includegraphics[width=7cm]figs/blackbox.pdf\nWe also test our method on 122 black-box problems from SRbench. Unlike ground-truth problems, in these problems, there are no underlying formulas describing the patterns in data. Ther... | [
{
"id": "5cuIwU35z1",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposes a search objective based on the principle of minimum description length, which aims to streamline and enhance symbolic regression tasks. The a... | {
"rating": "3;5;8",
"rating_avg": 5.333333333333333,
"confidence": "4;3;4",
"confidence_avg": 3.6666666666666665,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;3;4",
"contribution_avg": 3,
"presentation": "3;2;3",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.431592"
} | {
"id": "LAMpA3pqL0",
"metareview": "This paper proposes a novel method for symbolic regression, namely discovering the symbolic formula fitting data. Unlike existing methods where the difference between the symbolic function and the actual data is minimized during the search steps, this paper proposed to minimize ... | {
"decision": "Accept (Poster)"
} |
lja4JMesmC | 2410.06456v1 | From Generalist to Specialist: Adapting Vision Language Models via Task-Specific Visual Instruction Tuning | {
"content": "## Abstract\n\nAbstract Large vision language models (VLMs) combine large language models with vision encoders, demonstrating promise across various tasks. However, they often underperform in task-specific applications due to domain gaps between pre-training and fine-tuning.\nWe introduce VITask, a nove... | [
{
"id": "Ipy1NNS132",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposes a VITask framework that enhances the adaptability of Vision Language Models (VLMs) for specific tasks by integrating Task-Specific Models (TSM... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "4;3;4;4",
"confidence_avg": 3.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;2;4;2",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.432125"
} | {
"id": "wHFWoDLciU",
"metareview": "The paper introduces VITask, a framework for adapting vision-language models (VLMs) to task-specific applications using exemplar prompting, response distribution alignment, and contrastive response tuning. While reviewers appreciated the clarity of the methodology and its strong... | {
"decision": "Reject"
} |
lk2Qk5xjeu | 2409.02772v1 | Unifying Causal Representation Learning with the Invariance Principle | {
"content": "## Abstract\n\nAbstract Causal representation learning aims at recovering latent causal variables from high-dimensional observations to solve causal downstream tasks, such as predicting the effect of new interventions or more robust classification.\nA plethora of methods have been developed, each tackli... | [
{
"id": "qy2djraAH7",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 4,
"presentation": 2,
"summary": "This work provides a unifying theory for multiple causal representation learning works by using the fact that all these works are learning some form of invariance... | {
"rating": "5;5;8;8",
"rating_avg": 6.5,
"confidence": "4;4;2;3",
"confidence_avg": 3.25,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "4;4;3;3",
"contribution_avg": 3.5,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.432952"
} | {
"id": "htajlZga3X",
"metareview": "Invariances are at the core of causal modeling, but also at a large variety of approaches for robust modeling that incorporate theoretical rather than purely data-driven approaches for learning. Connections about the different ways of invariant representations from causal and no... | {
"decision": "Accept (Poster)"
} |
lmKJ1b6PaL | 2405.16507v3 | Causal Concept Graph Models: Beyond Causal Opacity in Deep Learning | {
"content": "## Abstract\n\nAbstract Causal opacity denotes the difficulty in understanding the “hidden” causal structure underlying the decisions of deep neural network (DNN) models.\nThis leads to the inability to rely on and verify state-of-the-art DNN-based systems, especially in high-stakes scenarios.\nFor this... | [
{
"id": "OQ8awPOMK3",
"initial_rating": 8,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper presents Causal Concept Graph Models (Causal CGMs), a concept-based architecture whose decision-making process is causally transparent by design. It av... | {
"rating": "3;5;5;8;8",
"rating_avg": 5.8,
"confidence": "3;3;4;2;2",
"confidence_avg": 2.8,
"soundness": "1;4;2;3;3",
"soundness_avg": 2.6,
"contribution": "2;1;2;3;3",
"contribution_avg": 2.2,
"presentation": "2;3;1;3;3",
"presentation_avg": 2.4
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.434049"
} | {
"id": "dyZiGeEocC",
"metareview": "The paper shows how to construct causal concept bottleneck models (CCBMs), i.e., interpretable models that aim to reveal the underlying decision process of deep learning models. Overall, the reviewers lean towards acceptance, and I agree. While the datasets considered are rathe... | {
"decision": "Accept (Poster)"
} |
ln2k0PqVQA | 2410.23022v1 | Online Intrinsic Rewards for Decision Making Agents from Large Language Model Feedback | {
"content": "## Abstract\n\nAbstract Automatically synthesizing dense rewards from natural language descriptions is a promising paradigm in reinforcement learning (RL), with applications to sparse reward problems, open-ended exploration, and hierarchical skill design. Recent works have made promising steps by exploi... | [
{
"id": "tEAcS0rGZ6",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This work presents a distributed system and an online learning algorithm for developing a reinforcement learning policy that learns from an intrinsic reward funct... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "4;3;4;3",
"confidence_avg": 3.5,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "1;2;2;2",
"contribution_avg": 1.75,
"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": "2026-01-14T22:16:03.435211"
} | {
"id": "Sln4xhVIFS",
"metareview": "This submission presents a framework for incorporating LLMs into an online RL setup by learning an intrinsic reward function concurrently with the policy. The idea is to asynchronously gather LLM annotations and distill them into a learned reward model, enabling sparse-reward pr... | {
"decision": "Reject"
} |
lo3nlFHOft | 2410.11998v1 | From Promise to Practice: Realizing High-performance Decentralized Training | {
"content": "## Abstract\n\nAbstract Decentralized training of deep neural networks has attracted significant attention for its theoretically superior scalability over synchronous data-parallel methods like All-Reduce. However, realizing this potential in multi-node training is challenging due to the complex design ... | [
{
"id": "k3fSRcrSn2",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The paper investigates decentralized training of DNNs. This used to be a fairly hot topic a few years ago (e.g. Lian et al 2017 and follow-up work), but has to so... | {
"rating": "5;5;8",
"rating_avg": 6,
"confidence": "5;4;3",
"confidence_avg": 4,
"soundness": "3;2;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.436429"
} | {
"id": "TPZGgM7g1r",
"metareview": "The paper studies decentralized training in the classic data parallel setting. It gives an analytical model to understand under what conditions decentralized training can bring gains. It also gives a relevant decentralized version of Adam, with a novel convergence analysis. Fina... | {
"decision": "Accept (Poster)"
} |
lpwS5T1jFb | 2410.08007v1 | Time Can Invalidate Algorithmic Recourse | {
"content": "## Abstract\n\nAbstract Algorithmic Recourse (AR) aims to provide users with actionable steps to overturn unfavourable decisions made by machine learning predictors.\nHowever, these actions often take time to implement ( e.g. , getting a degree can take years), and their effects may vary as the world ev... | [
{
"id": "HyXO7cAs1G",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper raises an interesting point: because it takes time to implement the recourse, the distributions of the features may shift. Thus, in the future, after th... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "4;4;3;4",
"confidence_avg": 3.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"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": "2026-01-14T22:16:03.437915"
} | {
"id": "dZWKHvqC7j",
"metareview": "I read the paper and all review/rebuttals.\n\nThe paper presents a way of setting up algorithmic recourse (AR) in the context of longitudinal causal models. This makes a lot of sense, as recourses take time to implement, and ultimately are meant to modify the state of the system... | {
"decision": "Reject"
} |
lqHv6dxBkj | 2405.16325v2 | SLoPe: Double-Pruned Sparse Plus Lazy Low-Rank Adapter Pretraining of LLMs | {
"content": "## Abstract\n\nAbstract We propose SLoPe , a Double-Pruned S parse Plus Lazy Lo w-rank Adapter P r e training method for LLMs that improves the accuracy of sparse LLMs while accelerating their pretraining and inference and reducing their memory footprint.\nSparse pretraining of LLMs reduces the accuracy... | [
{
"id": "f2zY0XLoGq",
"initial_rating": 6,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces SLOPE, a novel pretraining method aimed at improving the efficiency of large language models (LLMs) by combining sparsity and low-rank appro... | {
"rating": "5;6;6",
"rating_avg": 5.666666666666667,
"confidence": "4;4;2",
"confidence_avg": 3.3333333333333335,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;3;3",
"contribution_avg": 2.6666666666666665,
"presentation": "3;2;3",
"presentation_avg": 2.66666666666666... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.438858"
} | {
"id": "S1UgswzXPZ",
"metareview": "This paper introduces a novel pruning technique based on double-direction masks, which is further combined with LoRA in the final stages, resulting in the Double-Pruned Sparse Plus Lazy Low-rank Adapter Pre-training (SLoPE). The reviewers recognized the solid technical contribut... | {
"decision": "Accept (Poster)"
} |
lvhEptUoFF | 2410.01736v1 | Recursive Abstractive Processing for Retrieval in Dynamic Datasets | {
"content": "## Abstract\n\nAbstract Recent retrieval-augmented models enhance basic methods by building a hierarchical structure over retrieved text chunks through recursive embedding, clustering, and summarization.\nThe most relevant information is then retrieved from both the original text and generated summaries... | [
{
"id": "dmLorgSV06",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper addresses the dynamic document challenge problem in RAG for recursive abstractive indexing of a corpus. Specifically, to efficiently maintain the recur... | {
"rating": "3;3;3",
"rating_avg": 3,
"confidence": "3;3;4",
"confidence_avg": 3.3333333333333335,
"soundness": "2;2;2",
"soundness_avg": 2,
"contribution": "1;1;2",
"contribution_avg": 1.3333333333333333,
"presentation": "2;1;2",
"presentation_avg": 1.6666666666666667
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.440344"
} | {
"id": "7JMjs18jQk",
"metareview": "This paper addresses the challenge of updating retrieval corpora for RAG systems when new documents are available . The authors propose adRAP for efficiently updating the recursive abstractive indexing tree, and postQFRAP, a post-retrieval algorithm that builds a query-focused... | {
"decision": "Reject"
} |
lvw3UgeVxS | 2305.14749v6 | gRNAde: Geometric Deep Learning for 3D RNA inverse design | {
"content": "## Abstract\n\nAbstract Computational RNA design tasks are often posed as inverse problems, where sequences are designed based on adopting a single desired secondary structure without considering 3D geometry and conformational diversity.\nWe introduce gRNAde , a g eometric RNA de sign pipeline operating... | [
{
"id": "3L0lQ64YKy",
"initial_rating": 6,
"confidence": 3,
"soundness": 4,
"contribution": 4,
"presentation": 4,
"summary": "The manuscript presents gRNAde, an approach for RNA inverse design conditioned on the 3D structure of RNA (conformational ensemble). The method uses geometric dee... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "4;3;3;2",
"confidence_avg": 3,
"soundness": "4;3;4;4",
"soundness_avg": 3.75,
"contribution": "3;2;4;4",
"contribution_avg": 3.25,
"presentation": "1;3;4;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.441080"
} | {
"id": "Fybt4f7fOd",
"metareview": "This paper introduces gRNAde, a method for RNA sequence inverse design given 3D RNA structure information. \n\nAll reviewers agree that this is a solid paper with strong results. Similar techniques exist for general protein design (e.g. ProteinMPNN), but the paper adds sufficien... | {
"decision": "Accept (Spotlight)"
} |
lwTTZkDWoT | 2406.01514v3 | Decoupled Alignment for Robust Plug-and-Play Adaptation | {
"content": "### 1 Introduction\n\nWe introduce Decoupled Alignment for Robust Plug-and-Play Adaptation (termed Dapa), a low-resource safety enhancement method for aligning large language models (LLMs).\nDapa aligns unaligned Large language models (LLMs) with ethical guidelines even without supervised fine-tuning (S... | [
{
"id": "a9Ln5vSpnP",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 1,
"summary": "This paper proposes an interesting idea of isolating the weights of LLMs that impact the safety alignment the most. It then applies these weights to unaligned mod... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "4;4;3;3",
"confidence_avg": 3.5,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "3;1;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.441743"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
lzdFImKK8w | 2410.09543v1 | Boltzmann-Aligned Inverse Folding Model as a Predictor of Mutational Effects on Protein-Protein Interactions | {
"content": "## Abstract\n\nAbstract Predicting the change in binding free energy ( Δ Δ G Δ Δ 𝐺 \\Delta\\Delta G roman_Δ roman_Δ italic_G ) is crucial for understanding and modulating protein-protein interactions, which are critical in drug design.\nDue to the scarcity of experimental Δ Δ G Δ Δ 𝐺 \\Delta\\... | [
{
"id": "FCU5M5FcEA",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper employs pre-trained inverse folding models based on the Boltzmann distribution and thermodynamic cycles to associate energy with log-likelihood from in... | {
"rating": "6;6;6;8",
"rating_avg": 6.5,
"confidence": "4;3;3;3",
"confidence_avg": 3.25,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "4;3;3;3",
"contribution_avg": 3.25,
"presentation": "3;3;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.442288"
} | {
"id": "jNyK55JpE3",
"metareview": "The paper considers protein-protein interactions and proposes Boltzmann Alignment to predict the change in binding free energy resulting from mutation. The approach make use of physical inductive bias through Boltzmann distribution and thermodynamic cycle to link pre-trained inv... | {
"decision": "Accept (Spotlight)"
} |
m08aK3xxdJ | 2410.12261v1 | CATCH: Channel-Aware Multivariate Time Series Anomaly Detection via Frequency Patching | {
"content": "## Abstract\n\nAbstract Anomaly detection in multivariate time series is challenging as heterogeneous subsequence anomalies may occur. Reconstruction-based methods, which focus on learning nomral patterns in the frequency domain to detect diverse abnormal subsequences, achieve promising resutls, while s... | [
{
"id": "wPZrgAJIE1",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposed a method called CATCH \n for Multivariate Time Series Anomaly Detection.",
"strengths": "1. The paper is well structured.\n\n2. The archite... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;3;3;2",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.443035"
} | {
"id": "OrNMoZtaMk",
"metareview": "The paper tackles a multivariate time series anomaly detection (MTSAD) problem and introduces a channel-dependent approach to detect diverse point and subsequence anomalies in MTSAD. The method is evaluated on 11 MTSAD datasets.\n\nThe strengths of the work can be summarized as ... | {
"decision": "Accept (Poster)"
} |
m2gVfgWYDO | 2410.02094v2 | Tracking objects that change in appearance with phase synchrony | {
"content": "## Abstract\n\nAbstract Objects we encounter often change appearance as we interact with them. Changes in illumination (shadows), object pose, or movement of nonrigid objects can drastically alter available image features. How do biological visual systems track objects as they change? It may involve spe... | [
{
"id": "CLB8TGD2LH",
"initial_rating": 6,
"confidence": 4,
"soundness": 4,
"contribution": 4,
"presentation": 4,
"summary": "This paper presents CV-RNN, a deep learning circuit designed to replicate the biological visual system's ability to track objects with changing appearances by lev... | {
"rating": "3;6;8",
"rating_avg": 5.666666666666667,
"confidence": "4;4;4",
"confidence_avg": 4,
"soundness": "2;4;3",
"soundness_avg": 3,
"contribution": "2;4;3",
"contribution_avg": 3,
"presentation": "2;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.444018"
} | {
"id": "DYQYgI3t5z",
"metareview": "This paper introduces a complex valued RNN designed to replicate the biological visual system's ability to track objects with changing appearances by leveraging neural synchrony. The authors also introduce a new “Feature Tracker” benchmark and show that conventional deep learni... | {
"decision": "Accept (Poster)"
} |
m2nmp8P5in | 2404.18400v2 | LLM-SR: Scientific Equation Discovery via Programming with Large Language Models | {
"content": "## Abstract\n\nAbstract Mathematical equations have been unreasonably effective in describing complex natural phenomena across various scientific disciplines. However, discovering such insightful equations from data presents significant challenges due to the necessity of navigating extremely high-dimens... | [
{
"id": "jlNTNzasbs",
"initial_rating": 8,
"confidence": 3,
"soundness": 4,
"contribution": 4,
"presentation": 4,
"summary": "This paper introduces LLM-SR, a new approach to discovering mathematical equations from data using LLMs' scientific knowledge and code generation. LLM-SR treats e... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "5;4;3;3",
"confidence_avg": 3.75,
"soundness": "2;3;3;4",
"soundness_avg": 3,
"contribution": "3;2;3;4",
"contribution_avg": 3,
"presentation": "3;3;3;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.445093"
} | {
"id": "KrlF09suDl",
"metareview": "This paper proposes LLM-SR, a framework that leverages large language models for symbolic regression by integrating program synthesis, numerical optimization, and evolutionary search to discover accurate and generalizable scientific equations. Most of the reviewers agreed that t... | {
"decision": "Accept (Oral)"
} |
m9RNBZewW2 | 2410.04161v1 | Overcoming False Illusions in Real-World Face Restoration with Multi-Modal Guided Diffusion Model | {
"content": "## Abstract\n\nAbstract We introduce a novel Multi-modal Guided Real-World Face Restoration (MGFR) technique designed to improve the quality of facial image restoration from low-quality inputs.\nLeveraging a blend of attribute text prompts, high-quality reference images, and identity information, MGFR c... | [
{
"id": "BjAwQr579P",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper propose to utilize diffusion prior, as long as multi-modality input to perform real world face restoration. The method can receive text or reference ima... | {
"rating": "6;6;6",
"rating_avg": 6,
"confidence": "4;5;4",
"confidence_avg": 4.333333333333333,
"soundness": "3;4;3",
"soundness_avg": 3.3333333333333335,
"contribution": "3;3;2",
"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": "2026-01-14T22:16:03.446174"
} | {
"id": "SZEDsjXHDK",
"metareview": "The paper presents the Multi-modal guided face restoration method, which significantly improves face restoration from low-quality inputs using attribute prompts and reference images. Using the Reface-HQ dataset of over 23,000 images, it enhances visual quality and identity accur... | {
"decision": "Accept (Spotlight)"
} |
mAmCdASmJ5 | 2410.06993v1 | Efficient Distribution Matching of Representations via Noise-Injected Deep InfoMax | {
"content": "## Abstract\n\nAbstract Deep InfoMax (DIM) is a well-established method for self-supervised representation learning (SSRL) based on maximization of the mutual information between the input and the output of a deep neural network encoder.\nDespite the DIM and contrastive SSRL in general being well-explor... | [
{
"id": "qFWBJcy2Fc",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This work introduces noise injection to shape a self-supervised learning represenation as Gaussian or Uniformed representation. In short, by injecting a Guassian... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "3;3;4;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.447116"
} | {
"id": "RkZHIqCvzS",
"metareview": "This paper introduces a method for enforcing representations learnt with self-supervised learning (specifically Deep InfoMax) to follow a specific distribution by normalizing and injecting appropriate noise into encoder outputs. The method is simple yet novel, and is theoretical... | {
"decision": "Accept (Poster)"
} |
mBXLtNKpeQ | 2410.04543v1 | Pullback Flow Matching on Data Manifolds | {
"content": "## Abstract\n\nAbstract We propose \\ac PFM, a novel framework for generative modeling on data manifolds. Unlike existing methods that assume or learn restrictive closed-form manifold mappings for training \\ac RFM models, \\ac PFM leverages pullback geometry and isometric learning to preserve the under... | [
{
"id": "EoPO1dfVHw",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The paper introduces Pullback Flow Matching (PFM), a novel generative modeling framework designed to operate on data manifolds, which uses pullback geometry and i... | {
"rating": "3;5;5",
"rating_avg": 4.333333333333333,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "3;2;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;2;2",
"contribution_avg": 2,
"presentation": "2;2;3",
"presentation_avg": 2.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.447907"
} | {
"id": "Nl5T1pudPS",
"metareview": "This paper presents Pullback Flow Matching (PFM), a generative modeling framework for data manifolds that uses pullback geometry and isometric learning to preserve manifold geometry. By enhancing isometric learning with Neural ODEs and a scalable training objective, PFM is argue... | {
"decision": "Reject"
} |
mDKxlfraAn | 2410.05470v1 | Image Watermarks are Removable using Controllable Regeneration from Clean Noise | {
"content": "## Abstract\n\nAbstract Image watermark techniques provide an effective way to assert ownership, deter misuse, and trace content sources, which has become increasingly essential in the era of large generative models. A critical attribute of watermark techniques is their robustness against various manipu... | [
{
"id": "aVF6EtAmgf",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "1. The paper introduces CtrlRegen, a new method for removing image watermarks using controllable diffusion models. The method starts with clean Gaussian noise and... | {
"rating": "5;5;6;6;6",
"rating_avg": 5.6,
"confidence": "4;4;4;5;4",
"confidence_avg": 4.2,
"soundness": "3;3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;2;2;2",
"contribution_avg": 2,
"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": "2026-01-14T22:16:03.448667"
} | {
"id": "Zj0DyXCFmc",
"metareview": "This paper works on watermark removal. To achieve this, authors regenerates the watermarked images starting from a clean Gaussian noise via a controllable diffusion model. From experimental results, authors showed that proposal method is better than previous work. \n\nReviewers ... | {
"decision": "Accept (Poster)"
} |
mDvL3wcmms | 2410.03505v1 | Classification-denoising networks | {
"content": "## Abstract\n\nAbstract Image classification and denoising suffer from complementary issues of lack of robustness or partially ignoring conditioning information. We argue that they can be alleviated by unifying both tasks through a model of the joint probability of (noisy) images and class labels. Class... | [
{
"id": "xZ3Tkz19HH",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This work proposes a joint framework for classifying and denoising and also proposes a new network called GradResNet. This single network was trained for both den... | {
"rating": "1;3;5;5",
"rating_avg": 3.5,
"confidence": "5;3;3;2",
"confidence_avg": 3.25,
"soundness": "3;2;3;2",
"soundness_avg": 2.5,
"contribution": "1;1;2;2",
"contribution_avg": 1.5,
"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": "2026-01-14T22:16:03.449592"
} | {
"id": "RdUw9MWkkY",
"metareview": "Reviewers agree that the paper should not be accepted in its current form, due to underperformance wrt. prior work.\nAuthors rebuttal did not convince the reviewers of the papers limitation. Therefore the paper is recommended for rejection.",
"additional_comments": "Authors di... | {
"decision": "Reject"
} |
mEACsjW10N | 2409.17692v2 | MIO: A Foundation Model on Multimodal Tokens | {
"content": "## Abstract\n\nAbstract In this paper, we introduce MIO, a novel foundation model built on multimodal tokens, capable of understanding and generating speech, text, images, and videos in an end-to-end, autoregressive manner. While the emergence of large language models (LLMs) and multimodal large languag... | [
{
"id": "akFXzXVEVH",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper proposes a discrete token-based foundation model to unify understanding and generation capability in four modalities: speech, text, images, and videos ... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;3;2;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.450504"
} | {
"id": "eFnowFPZNl",
"metareview": "This paper introduces MIO, a new multimodal foundation model designed for understanding and generating content across speech, text, images, and videos. MIO uses a discrete token-based approach (Discrete-In-Discrete-Out or DIDO). The model employees a multi-stage training recipe ... | {
"decision": "Reject"
} |
mEBSeSk49H | 2403.15146v1 | On the Convergence of Adam under Non-uniform Smoothness: Separability from SGDM and Beyond | {
"content": "## Abstract\n\nAbstract This paper aims to clearly distinguish between Stochastic Gradient Descent with Momentum (SGDM) and Adam in terms of their convergence rates. We demonstrate that Adam achieves a faster convergence compared to SGDM under the condition of non-uniformly bounded smoothness. Our findi... | [
{
"id": "lW1RQtyJjX",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper provides a comparative analysis of Stochastic Gradient Descent with Momentum (SGDM) and simplified Adam (considering the same step-size for each coordi... | {
"rating": "1;5;5;8",
"rating_avg": 4.75,
"confidence": "4;4;3;2",
"confidence_avg": 3.25,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "1;3;2;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": "2026-01-14T22:16:03.451666"
} | {
"id": "22zQ4s2toT",
"metareview": "The paper presents a convergence analysis comparing Adam and SGDM under non-uniform smoothness assumptions in both deterministic and stochastic settings. The study introduces stopping-time techniques to show that Adam achieves convergence rates that align with lower bounds for f... | {
"decision": "Reject"
} |
mEpqHvbD2h | 2409.00588v2 | Diffusion Policy Policy Optimization | {
"content": "## Abstract\n\nAbstract We introduce Diffusion Policy Policy Optimization , DPPO , an algorithmic framework including best practices for fine-tuning diffusion-based policies (e.g. Diffusion Policy [ 20 ] ) in continuous control and robot learning tasks using the policy gradient (PG) method from reinforc... | [
{
"id": "GumMnoRIvt",
"initial_rating": 6,
"confidence": 4,
"soundness": 4,
"contribution": 2,
"presentation": 4,
"summary": "The authors propose a framework for finetuning a pretrained diffusion policy using on-policy RL, specifically PPO, and show that their method improves performance... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "2;3;2;4",
"soundness_avg": 2.75,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "1;3;3;4",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.453312"
} | {
"id": "yTGGBnmVEY",
"metareview": "This paper introduces a technique for fine-tuning diffusion policies with a policy gradient method. Previously, policy gradient updates of diffusion policies were conjectured to have training instability. With the new technique proposed, the authors show that the resulting algor... | {
"decision": "Accept (Poster)"
} |
mIjblC9hfm | 2407.09709v1 | GOFA: A Generative One-For-All Model for Joint Graph Language Modeling | {
"content": "## Abstract\n\nAbstract Foundation models, such as Large Language Models (LLMs) or Large Vision Models (LVMs), have emerged as one of the most powerful tools in the respective fields. However, unlike text and image data, graph data do not have a definitive structure, posing great challenges to developin... | [
{
"id": "SQ2Iv0z6gk",
"initial_rating": 8,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes a new joint graph and language model that can perform different graph and language tasks. The model is a combination of an LLM compressor, GNN ... | {
"rating": "5;5;6;8",
"rating_avg": 6,
"confidence": "4;4;4;2",
"confidence_avg": 3.5,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "2;4;4;3",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.454294"
} | {
"id": "VAk61AJUEt",
"metareview": "This paper presents GOFA, a generative model for joint graph and language modeling that interleaves GNN layers with a frozen LLM to address graph-structured tasks effectively. Reviewers commended the paper for its relevance, strong experimental results, and well-structured prese... | {
"decision": "Accept (Poster)"
} |
mKM9uoKSBN | 2410.14730v1 | On the Relation Between Linear Diffusion and Power Iteration | {
"content": "## Abstract\n\nAbstract Recently, diffusion models have gained popularity due to their impressive generative abilities. These models learn the implicit distribution given by the training dataset, and sample new data by transforming random noise through the reverse process, which can be thought of as gra... | [
{
"id": "dPT56P1zNt",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper studies the generation process of diffusion models. Focusing on linear models where the diffusion process is similar to performing noisy PCA, the autho... | {
"rating": "3;3;6",
"rating_avg": 4,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "1;2;3",
"soundness_avg": 2,
"contribution": "1;2;3",
"contribution_avg": 2,
"presentation": "2;2;3",
"presentation_avg": 2.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.455018"
} | {
"id": "s5gI1W5Xr9",
"metareview": "This paper presents a theoretical study on diffusion models. The main result is the establishment of a connection between power iteration and linear diffusion. It is claimed that this result can provide insight to the theoretical understanding of the generation process of diffus... | {
"decision": "Reject"
} |
mLxxv5gts0 | 2410.10180v1 | Gaussian Mixture Vector Quantization with Aggregated Categorical Posterior | {
"content": "## Abstract\n\nAbstract The vector quantization is a widely used method to map continuous representation to discrete space and has important application in tokenization for generative mode, bottlenecking information and many other tasks in machine learning.\nVector Quantized Variational Autoencoder (VQ-... | [
{
"id": "VSKS6q6r21",
"initial_rating": 1,
"confidence": 3,
"soundness": 1,
"contribution": 2,
"presentation": 2,
"summary": "The present paper introduces an idea to introduce a mixture of gaussians as a prior in a VQ-VAE. Furthermore, they introduce what is called Aggregated Categorical... | {
"rating": "1;3;5;5;5",
"rating_avg": 3.8,
"confidence": "3;5;3;4;4",
"confidence_avg": 3.8,
"soundness": "1;1;2;2;2",
"soundness_avg": 1.6,
"contribution": "2;2;2;2;2",
"contribution_avg": 2,
"presentation": "2;3;3;2;3",
"presentation_avg": 2.6
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.455661"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
mMPMHWOdOy | 2308.09583v1 | WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-Instruct | {
"content": "## Abstract\n\nAbstract Large language models (LLMs), such as GPT-4, have shown remarkable performance in natural language processing (NLP) tasks, including challenging mathematical reasoning. However, most existing open-source models are only pre-trained on large-scale internet data and without math-re... | [
{
"id": "QRJ669FbBC",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 4,
"presentation": 2,
"summary": "The paper takes Evol-Instruct (from the WizardLM paper) and extends it to the math domain, while at the same time integrating process reward models into the train... | {
"rating": "5;8;8;8",
"rating_avg": 7.25,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "2;3;4;3",
"soundness_avg": 3,
"contribution": "2;4;4;4",
"contribution_avg": 3.5,
"presentation": "3;3;3;2",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.456338"
} | {
"id": "gFK2ZLE3fu",
"metareview": "This paper takes the Evol-Instruct method invented in the WizardLM paper and applied that to the math domain to create a strong math SFT dataset. Then the authors also perform reinforcement learning to further boost the performance. \n\nThis paper receives a high score of 8, 8, ... | {
"decision": "Accept (Oral)"
} |
mMPaQzgzAN | 2407.14435v3 | Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoders | {
"content": "## Abstract\n\nAbstract Sparse autoencoders (SAEs) are a promising unsupervised approach for identifying causally relevant and interpretable linear features in a language model’s (LM) activations.\nTo be useful for downstream tasks, SAEs need to decompose LM activations faithfully; yet to be interpretab... | [
{
"id": "RqqoJ7plxJ",
"initial_rating": 3,
"confidence": 4,
"soundness": 1,
"contribution": 1,
"presentation": 2,
"summary": "For LM activation decomposition, the paper proposes to use JumpReLU activation and $L^0$ norm instead of $L^1$ norm for sparse autoencoders. The proposed method o... | {
"rating": "3;3;3;8",
"rating_avg": 4.25,
"confidence": "2;3;4;3",
"confidence_avg": 3,
"soundness": "2;3;1;4",
"soundness_avg": 2.5,
"contribution": "2;2;1;3",
"contribution_avg": 2,
"presentation": "3;3;2;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.457033"
} | {
"id": "Aw4CKmIj4P",
"metareview": "This work introduces JumpReLU SAEs, which the authors claim achieve state-of-the-art reconstruction fidelity at a given sparsity level on Gemma 2 9B activations, surpassing recent advances such as Gated and TopK SAEs. The authors also claim that these gains do not compromise int... | {
"decision": "Reject"
} |
mMXdHyBcHh | 2409.02897v3 | LongCite: Enabling LLMs to Generate Fine-grained Citations in Long-context QA | {
"content": "## Abstract\n\nAbstract Though current long-context large language models (LLMs) have demonstrated impressive capacities in answering user questions based on extensive text, the lack of citations in their responses makes user verification difficult, leading to concerns about their trustworthiness due to... | [
{
"id": "X8MRL1bYlw",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper introduces LongBench-Cite, a Long-Context QA benchmark that requires models to add citations in their responses. Different from other similar works, th... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "2;4;5;4",
"confidence_avg": 3.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "2;3;4;2",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.457731"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
mMhZS7qt0U | 2408.09730v1 | Fragment and Geometry Aware Tokenization of Molecules for Structure-Based Drug Design Using Language Models | {
"content": "## Abstract\n\nAbstract Structure-based drug design (SBDD) is crucial for developing specific and effective therapeutics against protein targets but remains challenging due to complex protein-ligand interactions and vast chemical space. Although language models (LMs) have excelled in natural language pr... | [
{
"id": "VPfu1TSqUv",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper presents Frag2Seq, a method for structure-based drug design using language models (LMs) to generate drug-like molecules based on 3D fragments. By transf... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "5;4;3;3",
"confidence_avg": 3.75,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"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": "2026-01-14T22:16:03.458582"
} | {
"id": "vuyui9z3V8",
"metareview": "This paper proposes a method to generate molecules in SE(3) equivariant way using SE(3) invariant tokenization of molecular fragments. The tokenization is based on canonicalized ordering of atoms (and fragments) and serialization based on fragments and relative rotations between... | {
"decision": "Accept (Poster)"
} |
mMmzHS28ht | 2408.11796v2 | LLM Pruning and Distillation in Practice | {
"content": "## Abstract\n\nAbstract Abstract: We present a comprehensive report on compressing the Llama 3.1 8B and Mistral NeMo 12B models to 4B and 8B parameters, respectively, using pruning and distillation [ 1 ] .\nWe explore two distinct pruning strategies: (1) depth pruning and (2) joint hidden/attention/MLP ... | [
{
"id": "W0yR4hCUs8",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper studies the combination of structural pruning and knowledge distillation to obtain compressed and performant language models with higher throughput. Th... | {
"rating": "3;5;6",
"rating_avg": 4.666666666666667,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;2;2",
"contribution_avg": 2,
"presentation": "1;2;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.459446"
} | {
"id": "ZT4vmNBIAg",
"metareview": "This paper tackles an increasingly relevant challenge in the field of large language models (LLMs): compressing models effectively when access to the original training data is restricted. The key contributions include a novel \"teacher correction\" phase to adapt the teacher mod... | {
"decision": "Reject"
} |
mOpNrrV2zH | 2406.10840v3 | CBGBench: Fill in the Blank of Protein-Molecule Complex Binding Graph | {
"content": "## Abstract\n\nAbstract Structure-based drug design (SBDD) aims to generate potential drugs that can bind to a target protein and is greatly expedited by the aid of AI techniques in generative models. However, a lack of systematic understanding persists due to the diverse settings, complex implementatio... | [
{
"id": "4vbjYIGfwh",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper proposes a benchmark for SBDD, including a unified framework of generative graph completion for multiple tasks in the field and a comprehensive evaluat... | {
"rating": "3;6;6;6;8",
"rating_avg": 5.8,
"confidence": "2;4;4;4;4",
"confidence_avg": 3.6,
"soundness": "3;3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3;3",
"contribution_avg": 2.8,
"presentation": "2;3;3;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.461152"
} | {
"id": "HbXgPhMabH",
"metareview": "The paper makes a solid contribution by introducing a novel benchmark for models applied to structure-based drug discovery. \n\nReviewers appreciated the scope of the evaluated methods, the design of the subtasks, and the shared codebase. All reviewers voted to accept the paper.... | {
"decision": "Accept (Spotlight)"
} |
mPMLZv4kSL | 2411.16686v1 | ProteinWeaver: A Divide-and-Assembly Approach for Protein Backbone Design | {
"content": "## Abstract\n\nAbstract Nature creates diverse proteins through a ‘divide and assembly’ strategy. Inspired by this idea, we introduce ProteinWeaver, a two-stage framework for protein backbone design. Our method first generates individual protein domains and then employs an SE(3) diffusion model to flexi... | [
{
"id": "cXlckLLEE2",
"initial_rating": 3,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper introduces ProteinWeaver, a two-stage framework for protein backbone design. In the first stage, individual protein domains are generated, and in the se... | {
"rating": "3;3;3;6;6",
"rating_avg": 4.2,
"confidence": "5;4;4;3;4",
"confidence_avg": 4,
"soundness": "2;3;2;3;3",
"soundness_avg": 2.6,
"contribution": "2;1;2;3;3",
"contribution_avg": 2.2,
"presentation": "3;2;2;2;2",
"presentation_avg": 2.2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Desk Rejected Submission",
"venueid": "ICLR.cc/2025/Conference/Desk_Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.462168"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
mPdmDYIQ7f | 2410.06153v1 | AgentSquare: Automatic LLM Agent Search in Modular Design Space | {
"content": "## Abstract\n\nAbstract Recent advancements in Large Language Models (LLMs) have led to a rapid growth of agentic systems capable of handling a wide range of complex tasks. However, current research largely relies on manual, task-specific design, limiting their adaptability to novel tasks. In this paper... | [
{
"id": "nRERgG9rqA",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces AgentSquare, a framework designed to automatically optimize LLM agent architectures within a modular design space. It proposes a novel approa... | {
"rating": "1;5;5;6",
"rating_avg": 4.25,
"confidence": "5;3;3;3",
"confidence_avg": 3.5,
"soundness": "1;2;3;3",
"soundness_avg": 2.25,
"contribution": "1;3;3;3",
"contribution_avg": 2.5,
"presentation": "1;2;3;2",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.462950"
} | {
"id": "cCweqLEqBB",
"metareview": "Wow, working through the reviewer discussion for this paper was a trip. As someone who has seen multiple cases of plagiarism accusations, both well-founded and unfounded, it was really interesting to see how this played out. As it turns out, I am happy with the outcome. It seems... | {
"decision": "Accept (Poster)"
} |
mSYX71lNAl | 2410.08455v1 | Why pre-training is beneficial for downstream classification tasks? | {
"content": "## Abstract\n\nAbstract Pre-training has exhibited notable benefits to downstream tasks by boosting accuracy and speeding up convergence, but the exact reasons for these benefits still remain unclear.\nTo this end, we propose to quantitatively and explicitly explain effects of pre-training on the downst... | [
{
"id": "BOmIxATNf7",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This study analyzes how pre-training a model helps the fine-tuning process through knowledge-based decomposition. The authors define knowledge as the intersection... | {
"rating": "3;3;5;5;6",
"rating_avg": 4.4,
"confidence": "3;5;4;4;3",
"confidence_avg": 3.8,
"soundness": "2;2;3;3;3",
"soundness_avg": 2.6,
"contribution": "2;1;3;3;3",
"contribution_avg": 2.4,
"presentation": "2;1;2;3;3",
"presentation_avg": 2.2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.463595"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
mTgMLy2iPt | 2301.13236v1 | Policy Gradient with Tree Expansion | {
"content": "## Abstract\n\nAbstract Despite the popularity of policy gradient methods, they are known to suffer from large variance and high sample complexity.\nTo mitigate this, we introduce SoftTreeMax – a generalization of softmax that takes planning into account. In SoftTreeMax , we extend the traditional logit... | [
{
"id": "RLK7gIjEKK",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces a novel method to reduce variance in Policy Gradient (PG) methods by integrating Tree Search (TS) with a differentiable parametric policy. By... | {
"rating": "3;3;6;8",
"rating_avg": 5,
"confidence": "2;3;4;3",
"confidence_avg": 3,
"soundness": "1;2;3;3",
"soundness_avg": 2.25,
"contribution": "3;2;2;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": "2026-01-14T22:16:03.464772"
} | {
"id": "UAygfnKjhM",
"metareview": "This paper proposed and analyzed a softmax-linear policy parameterization with an interesting feature design that depends on the behavioral policy that generate the data and a tree search approach. Two methods called C-Soft-TreeMax and E-Soft-TreeMax have been proposed. Though t... | {
"decision": "Reject"
} |
mUXdysoxEP | 2405.17816v1 | Pursuing Feature Separation based on Neural Collapse for Out-of-Distribution Detection | {
"content": "## Abstract\n\nAbstract In the open world, detecting out-of-distribution (OOD) data, whose labels are disjoint with those of in-distribution (ID) samples, is important for reliable deep neural networks (DNNs). To achieve better detection performance, one type of approach proposes to fine-tune the model ... | [
{
"id": "tq1c3m4jI4",
"initial_rating": 5,
"confidence": 2,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper proposes an OOD detection approach that enhances feature separation. The proposed method is based on Neural Collapse (NC), which suggests that features ... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "4;2;4;5",
"confidence_avg": 3.75,
"soundness": "2;3;3;4",
"soundness_avg": 3,
"contribution": "2;2;3;3",
"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": "2026-01-14T22:16:03.465429"
} | {
"id": "F0XnqBdrUk",
"metareview": "The authors propose to include two additional terms to standard OE (outlier exposure) loss function for improved out of distribution (OOD) detection-- the first term is to ensure extremely compact clustering of in-distribution samples while the second one ensures that the OOD fe... | {
"decision": "Accept (Poster)"
} |
mV6cO4mGjH | 2402.12519v2 | Dynamics Based Neural Encoding with Inter-Intra Region Connectivity | {
"content": "## Abstract\n\nAbstract Extensive literature has drawn comparisons between recordings of biological neurons in the brain and deep neural networks. This comparative analysis aims to advance and interpret deep neural networks and enhance our understanding of biological neural systems.\nHowever, previous w... | [
{
"id": "Uhipytf7hH",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This work provides a comprehensive study of predicting fMRI signals in visual cortical areas while humans view short video clips, by mapping from the activations ... | {
"rating": "1;5;5;8",
"rating_avg": 4.75,
"confidence": "5;5;3;3",
"confidence_avg": 4,
"soundness": "1;2;3;4",
"soundness_avg": 2.5,
"contribution": "1;2;3;4",
"contribution_avg": 2.5,
"presentation": "1;2;2;4",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.466149"
} | {
"id": "oE5BjR65T2",
"metareview": "The paper presents a comprehensive study comparing image and video understanding models to visual cortex recordings using video stimuli. The study reveals insights into how different specs of deep networks predict neural responses. Moreover, the paper introduces a novel neural e... | {
"decision": "Reject"
} |
mVExccNdtK | 2402.08653v4 | SAGMAN: Stability Analysis of Graph Neural Networks on the Manifolds | {
"content": "## Abstract\n\nAbstract Modern graph neural networks (GNNs) can be sensitive to changes in the input graph structure and node features, potentially resulting in unpredictable behavior and degraded performance. In this work, we introduce a spectral framework known as SAGMAN for examining the stability of... | [
{
"id": "xNqfQtWxRk",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposes a complicated framework called SAGMAN to measure the stability of trained GNN model. \nIt has three phases. 1. Construct the graph embedding b... | {
"rating": "3;3;3;8",
"rating_avg": 4.25,
"confidence": "3;4;3;4",
"confidence_avg": 3.5,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;3;2;3",
"presentation_avg": 2.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.467250"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
mXHTifc1Fn | 2402.15864v1 | E(3)-equivariant models cannot learn chirality: Field-based molecular generation | {
"content": "## Abstract\n\nAbstract This work introduces FMG, a field-based model for drug-like molecule generation. We show how the flexibility of this method provides crucial advantages over the prevalent, point-cloud based methods, and achieves competitive molecular stability generation. We tackle optical isomer... | [
{
"id": "bheRJgUlQe",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "This paper tackles the limitations of E(3)-invariant features in modeling chirality, as mirror symmetry is inherently implied in such features. The authors first ... | {
"rating": "3;6;8;8",
"rating_avg": 6.25,
"confidence": "4;3;4;3",
"confidence_avg": 3.5,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"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": "2026-01-14T22:16:03.468086"
} | {
"id": "yiOToRLZh3",
"metareview": "This paper proposes a CNN-based method for molecular generation, focusing on effectively capturing chirality which usually fails to be \n incorporated in E(3) invariant features. The method represents molecules as functions in 3D space. Each atom is represented as a Gaussian dis... | {
"decision": "Accept (Poster)"
} |
mb2rHLcKN5 | 2408.11172v1 | SubgoalXL: Subgoal-based Expert Learning for Theorem Proving | {
"content": "## Abstract\n\nAbstract Formal theorem proving, a field at the intersection of mathematics and computer science, has seen renewed interest with advancements in large language models (LLMs). This paper introduces 𝚂𝚞𝚋𝚐𝚘𝚊𝚕𝚇𝙻 𝚂𝚞𝚋𝚐𝚘𝚊𝚕𝚇𝙻 \\mathtt{S}\\mathtt{u}\\mathtt{b}\\mathtt{g}\\mathtt{o... | [
{
"id": "AAIODlHjjp",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper introduces SubgoalXL: a LLM-based system for formal theorem proving. The system comprises of three components:\n\n(1) formal statement generator,\\\n(2... | {
"rating": "3;3;3;6",
"rating_avg": 3.75,
"confidence": "5;4;3;3",
"confidence_avg": 3.75,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "1;2;2;4",
"contribution_avg": 2.25,
"presentation": "1;1;2;3",
"presentation_avg": 1.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.469081"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
meKEKDhdnx | 2407.00023v2 | Preble: Efficient Distributed Prompt Scheduling for LLM Serving | {
"content": "## Abstract\n\nAbstract Prompts to large language models (LLMs) have evolved beyond simple user questions. For LLMs to solve complex problems, today’s practices are to include domain-specific instructions, illustration of tool usages, and/or long context such as textbook chapters in prompts. As such, ma... | [
{
"id": "1P4EZ1H3Rs",
"initial_rating": 3,
"confidence": 3,
"soundness": 1,
"contribution": 2,
"presentation": 3,
"summary": "The paper introduces Preble, a distributed LLM serving system specifically designed to handle long-context prompt workloads. The authors propose a tree-based pref... | {
"rating": "3;3;6;8;8",
"rating_avg": 5.6,
"confidence": "4;3;3;3;4",
"confidence_avg": 3.4,
"soundness": "2;1;3;3;4",
"soundness_avg": 2.6,
"contribution": "1;2;3;3;3",
"contribution_avg": 2.4,
"presentation": "2;3;3;3;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.470125"
} | {
"id": "977dHlHSU3",
"metareview": "The paper proposes \"Preble,\" a distributed LLM serving platform optimizing prompt sharing and co-optimizing KV state reuse with computation load-balancing. The authors introduce hierarchical scheduling mechanisms and a novel algorithm, E2, which purportedly outperforms existin... | {
"decision": "Accept (Poster)"
} |
meOELl7HRf | 2410.12781v1 | Long-LRM: Long-sequence Large Reconstruction Model for Wide-coverage Gaussian Splats | {
"content": "## Abstract\n\nAbstract We propose Long-LRM, a generalizable 3D Gaussian reconstruction model that is capable of reconstructing a large scene from\na long sequence of input images. Specifically, our model can process 32 source images at 960 × \\times 540 resolution within only 1.3 seconds on a single A1... | [
{
"id": "zWnQ8CsFbn",
"initial_rating": 3,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces Long-LRM, a feed-forward model for large-scale 3D Gaussian reconstruction that can process 32 high-resolution (960×540) input images in just... | {
"rating": "3;5;5",
"rating_avg": 4.333333333333333,
"confidence": "5;4;5",
"confidence_avg": 4.666666666666667,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;2;2",
"contribution_avg": 2,
"presentation": "3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.470884"
} | {
"id": "gAc07oTitj",
"metareview": "This paper proposes a feed-forward model for large-scale 3D Gaussian reconstruction that can process 32 high-resolution (960×540) input images in 1.3 seconds on a single A100 GPU. The key innovation lies in its hybrid architecture combining Mamba2 and transformer blocks, along w... | {
"decision": "Reject"
} |
medKq3cONT | 2410.09560v1 | Towards Scalable Semantic Representation for Recommendation | {
"content": "## Abstract\n\nAbstract With recent advances in large language models (LLMs), there has been emerging numbers of research in developing Semantic IDs based on LLMs to enhance the performance of recommendation systems.\nHowever, the dimension of these embeddings needs to match that of the ID embedding in ... | [
{
"id": "2l85liw7pW",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper investigates the scalability of semantic representation based on LLM for recommendation. The authors introduce an innovative approach termed Mixture-of... | {
"rating": "3;5;6;6",
"rating_avg": 5,
"confidence": "5;3;4;3",
"confidence_avg": 3.75,
"soundness": "1;2;3;3",
"soundness_avg": 2.25,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "1;2;4;2",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.471459"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
mgjRSXGnF0 | 2411.06071v1 | GlocalCLIP: Object-agnostic Global-Local Prompt Learning for Zero-shot Anomaly Detection | {
"content": "## Abstract\n\nAbstract Zero-shot anomaly detection (ZSAD) is crucial for detecting abnormal patterns in target datasets without using training samples, specifically in scenarios where there are distributional differences between the target domain and training data or where data scarcity arises because ... | [
{
"id": "cJ2TFFSSfg",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces a novel method for zero-shot anomaly detection. GlocalCLIP separates global and local prompts and optimize them together, enabling detection... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "4;5;5;4",
"confidence_avg": 4.5,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;1;2;2",
"contribution_avg": 1.75,
"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": "2026-01-14T22:16:03.472090"
} | {
"id": "Uaanj1kx64",
"metareview": "The paper proposes a zero shot anomaly detection method using CLIP features. There are several such methods and the contributions on the method lie in using a dual-branch approach to separately model global and local anomaly semantics. These are then jointly optimized via globa... | {
"decision": "Reject"
} |
mhyl7HhNM5 | 2410.18889v1 | Are LLMs Better than Reported? Detecting Label Errors and Mitigating Their Effect on Model Performance | {
"content": "## Abstract\n\nAbstract NLP benchmarks rely on standardized datasets for training and evaluating models and are crucial for advancing the field. Traditionally, expert annotations ensure high-quality labels; however, the cost of expert annotation does not scale well with the growing demand for larger dat... | [
{
"id": "NdfbegfHsd",
"initial_rating": 5,
"confidence": 4,
"soundness": 1,
"contribution": 2,
"presentation": 4,
"summary": "The paper examines mislabeled instances in existing standard label-based datasets. The proposed method uses large language models (LLMs) to identify these instanc... | {
"rating": "5;5;8",
"rating_avg": 6,
"confidence": "4;4;4",
"confidence_avg": 4,
"soundness": "3;1;4",
"soundness_avg": 2.6666666666666665,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "3;4;4",
"presentation_avg": 3.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.473167"
} | {
"id": "UgQr4KAK5m",
"metareview": "This paper investigates the use of LLMs to identify mislabeled instances in NLP datasets, comparing LLM predictions, expert annotations, and crowd-sourced labels. The study highlights the prevalence of labeling errors (10–40%) in four binary classification datasets and demonstra... | {
"decision": "Reject"
} |
mjtCqmujYP | 2410.08067v1 | Reward-Augmented Data Enhances Direct Preference Alignment of LLMs | {
"content": "## Abstract\n\nAbstract Preference alignment in Large Language Models (LLMs) has significantly improved their ability to adhere to human instructions and intentions. However, existing direct alignment algorithms primarily focus on relative preferences and often overlook the qualitative aspects of respon... | [
{
"id": "1MwshwatBS",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "The paper introduces a novel method to enhance the direct preference alignment of Large Language Models (LLMs) by incorporating reward-augmented data. The authors... | {
"rating": "3;3;5;6;6",
"rating_avg": 4.6,
"confidence": "4;4;4;3;4",
"confidence_avg": 3.8,
"soundness": "2;3;3;3;2",
"soundness_avg": 2.6,
"contribution": "1;2;2;2;3",
"contribution_avg": 2,
"presentation": "2;3;2;3;2",
"presentation_avg": 2.4
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.473890"
} | {
"id": "8gRGQ3a7z1",
"metareview": "The paper introduces a novel data relabeling method for LLM preference alignment that conditions training on explicit reward scores rather than just binary preferences. This approach helps prevent \"unlearning\" of high-quality rejected responses and enables better generalizatio... | {
"decision": "Reject"
} |
mkE9Yx4wHY | 2410.13216v1 | Anchored Alignment for Self-Explanations Enhancement | {
"content": "## Abstract\n\nAbstract In this work, we introduce a methodology for alignment designed to enhance the ability of large language models (LLMs) to articulate their reasoning— self-explanation —even in the absence of annotated rationale explanations. Our alignment methodology comprises three key component... | [
{
"id": "DkNI6RVHsd",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper introduces a new self-explanation alignment method, called Alignment with Anchor Preference Pairs, to improve both exact answer prediction and explanat... | {
"rating": "3;5;6;8",
"rating_avg": 5.5,
"confidence": "5;3;4;4",
"confidence_avg": 4,
"soundness": "2;3;3;4",
"soundness_avg": 3,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;3;4;4",
"presentation_avg": 3.25
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.474571"
} | {
"id": "Auxieqd7Ij",
"metareview": "This paper presents a method for improving methods' self-explanation abilities. It produces sets of preference pairs for alignment to find data that will improve these abilities. This is done by separating the examples into several cases (consistently correct, consistently incor... | {
"decision": "Reject"
} |
mkNVPGpEPm | 2410.13866v1 | Associative memory and dead neurons | {
"content": "## Abstract\n\nAbstract In “Large Associative Memory Problem in Neurobiology and Machine Learning,” Dmitry Krotov and John Hopfield introduced a general technique for the systematic construction of neural ordinary differential equations with non-increasing energy or Lyapunov function. We study this ener... | [
{
"id": "XqwMeIwwK7",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The work investigates Lyapunov functions and stability in the Krotov and Hopfield models of associative memory. It is found that the conventional formulation suff... | {
"rating": "3;5;8",
"rating_avg": 5.333333333333333,
"confidence": "2;2;4",
"confidence_avg": 2.6666666666666665,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "3;3;3",
"contribution_avg": 3,
"presentation": "1;1;2",
"presentation_avg": 1.3333333333333333
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.475234"
} | {
"id": "7Viw59Gr40",
"metareview": "**Summary:**\nThis paper studies the Krotov-Hopfield (KH) model of associative memory (Krotov-Hopfield, 2020), and points out that the energy function of their model suffers from the problem of dead neurons, implying that the energy may have flat directions, which in turn may be... | {
"decision": "Accept (Poster)"
} |
mlCRJnETWz | 2407.20224v3 | Can Editing LLMs Inject Harm? | {
"content": "## Abstract\n\nAbstract Knowledge editing has been increasingly adopted to correct the false or outdated knowledge in Large Language Models (LLMs). Meanwhile, one critical but under-explored question is: can knowledge editing be used to inject harm into LLMs? In this paper, we propose to reformulate kno... | [
{
"id": "1nZPy5TzEn",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 1,
"summary": "The paper studies malicious knowledge editing in LLMs as a safety threat. In particular, it considers misinformation and bias. It finds both threats are very poss... | {
"rating": "3;3;5;5;6",
"rating_avg": 4.4,
"confidence": "3;3;3;5;3",
"confidence_avg": 3.4,
"soundness": "3;3;3;2;3",
"soundness_avg": 2.8,
"contribution": "2;2;2;2;2",
"contribution_avg": 2,
"presentation": "2;1;3;3;3",
"presentation_avg": 2.4
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.476070"
} | {
"id": "Qdr9d1b6j8",
"metareview": "This work proposes the editing attack, which reveals the risk of knowledge editing in injecting harmful information into LLMs and affecting their fairness and accuracy. The EDITATTACK dataset it constructs and the findings it provides are recognized by the reviewers. However, th... | {
"decision": "Reject"
} |
mn61GWpEiK | 2405.13915v1 | HeteGraph-Mamba: Heterogeneous Graph Learning via Selective State Space Model | {
"content": "## Abstract\n\nAbstract We propose a heterogeneous graph mamba network (HGMN) as the first exploration in leveraging the selective state space models (SSSMs) for heterogeneous graph learning. Compared with the literature, our HGMN overcomes two major challenges: (i) capturing long-range dependencies amo... | [
{
"id": "TNflXnJouY",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposes a graph mamba network as a graph model to learn node representations, aiming to incorporate long-range dependencies for heterogeneous graphs. ... | {
"rating": "3;3;3;5",
"rating_avg": 3.5,
"confidence": "5;4;3;4",
"confidence_avg": 4,
"soundness": "2;2;2;2",
"soundness_avg": 2,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "3;2;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": "2026-01-14T22:16:03.476922"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
mnna9LUg7P | 2410.13229v1 | Quamba: A Post-Training Quantization Recipe for Selective State Space Models | {
"content": "## Abstract\n\nAbstract State Space Models (SSMs) have emerged as an appealing alternative to Transformers for large language models, achieving state-of-the-art accuracy with constant memory complexity which allows for holding longer context lengths than attention-based networks.\nThe superior computati... | [
{
"id": "NxmpWR8SA1",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 4,
"presentation": 3,
"summary": "This work propose a PTQ solution for SSM model. Quamba notices the SSMs have highly sensitive feature maps within the selective scan mechanism (i.e., linear recur... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "5;4;3;3",
"confidence_avg": 3.75,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "3;2;3;4",
"contribution_avg": 3,
"presentation": "1;3;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.477675"
} | {
"id": "V4V35JZlNI",
"metareview": "This paper presents an imporant contribution to SSM quantization, enabling faster inference with minimal accuracy loss. The authors carefully addressed initial reviewer concerns about novelty and theoretical clarity through additional experiments and improved explanations. Their... | {
"decision": "Accept (Poster)"
} |
moWiYJuSGF | 2410.13232v1 | Web Agents with World Models: Learning and Leveraging Environment Dynamics in Web Navigation | {
"content": "## Abstract\n\nAbstract Large language models (LLMs) have recently gained much attention in building autonomous agents. However, performance of current LLM-based web agents in long-horizon tasks is far from optimal, often yielding errors such as repeatedly buying a non-refundable flight ticket. By contr... | [
{
"id": "lSOZnweVBa",
"initial_rating": 6,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "The paper proposes an approach for enhancing the performance of LLM-based web agents in long-horizon tasks. The authors introduce a World-Model-Augmented (WMA) we... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "3;3;2;2",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;1;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.478532"
} | {
"id": "WhVpuoi4Lk",
"metareview": "Summary:\n\nThis paper presents a World-model-augmented (WMA) web agent, which simulates the outcomes of the agent’s actions for better decision-making. To train a world model for web agents, the paper proposes a transition-focused observation abstraction, where the prediction o... | {
"decision": "Accept (Poster)"
} |
mqNKiEB6pd | 2407.03038v1 | Towards Federated RLHF with Aggregated Client Preference for LLMs | {
"content": "## Abstract\n\nAbstract Reinforcement learning with human feedback (RLHF) fine-tunes a pretrained large language model (LLM) using preference datasets, enabling the LLM to generate outputs that align with human preferences. Given the sensitive nature of these preference datasets held by various clients,... | [
{
"id": "VkiB6gdQOT",
"initial_rating": 6,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 2,
"summary": "Reinforcement learning with human feedback (RLHF) is a good way to fine-tune a pretrained large language model (LLM) to generate content aligned with human prefer... | {
"rating": "5;6;6;6",
"rating_avg": 5.75,
"confidence": "3;2;4;3",
"confidence_avg": 3,
"soundness": "3;3;3;2",
"soundness_avg": 2.75,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "4;3;3;2",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.479238"
} | {
"id": "ZXqID721bS",
"metareview": "Reinforcement learning with human feedback (RLHF) is an effective method for fine-tuning pretrained large language models (LLMs) to align their outputs with human preferences. This paper introduces a novel approach that leverages federated learning to aggregate diverse model inf... | {
"decision": "Accept (Poster)"
} |
mr2icR6dpD | 2406.05814v1 | TIGeR: Unifying Text-to-Image Generation and Retrieval with Large Multimodal Models | {
"content": "## Abstract\n\nAbstract How humans can efficiently and effectively acquire images has always been a perennial question.\nA typical solution is text-to-image retrieval from an existing database given the text query; however, the limited database typically lacks creativity.\nBy contrast, recent breakthrou... | [
{
"id": "wgjcT85SAW",
"initial_rating": 3,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper presents an approach to address the limitations of current text-to-image generation (T2I-G) and retrieval (T2I-R) systems. Particularly, it propose to ... | {
"rating": "3;6;6;6;8",
"rating_avg": 5.8,
"confidence": "3;2;4;3;4",
"confidence_avg": 3.2,
"soundness": "3;3;4;2;4",
"soundness_avg": 3.2,
"contribution": "2;3;3;3;4",
"contribution_avg": 3,
"presentation": "2;3;3;3;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.479930"
} | {
"id": "VcaQre5FtJ",
"metareview": "This paper presents a framework to unify text-to-image generation and retrieval by running the generation and retrieval branches, followed by a image selection process for the final output. \n\nAs reviewers pointed out, the major strengths of this paper are:\n1) The idea of unif... | {
"decision": "Accept (Poster)"
} |
mrNVOWlG25 | 2409.15219v1 | MotifDisco: Motif Causal Discovery For Time Series Motifs | {
"content": "## Abstract\n\nAbstract Many time series, particularly health data streams, can be best understood as a sequence of phenomenon or events, which we call motifs . A time series motif is a short trace segment which may implicitly capture an underlying phenomenon within the time series. Specifically, we foc... | [
{
"id": "vqZnzRosFc",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The paper proposes to learn granger causal graphs over time series segments. The segments are defined as motifs as they have specific characteristics. The segment... | {
"rating": "3;3;5",
"rating_avg": 3.6666666666666665,
"confidence": "3;4;4",
"confidence_avg": 3.6666666666666665,
"soundness": "3;2;2",
"soundness_avg": 2.3333333333333335,
"contribution": "1;2;2",
"contribution_avg": 1.6666666666666667,
"presentation": "2;3;3",
"presentation_avg": 2.6666666666666... | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.480648"
} | {
"id": "a8vyju6ruK",
"metareview": "MotifDisco is a framework that tries to find short segments representing underlying phenomena within time series. MotifDisco uses Granger Causality and Transfer Entropy to learn causal relationships by solving an unsupervised link prediction problem in a graph neural network. I ... | {
"decision": "Reject"
} |
msEr27EejF | 2403.03185v2 | Correlated Proxies: A New Definition and Improved Mitigation for Reward Hacking | {
"content": "## Abstract\n\nAbstract Because it is difficult to precisely specify complex objectives, reinforcement learning policies are often optimized using flawed proxy rewards that seem to capture the true objective. However, optimizing proxy rewards frequently leads to reward hacking : the optimized reward fun... | [
{
"id": "T1g1mByGam",
"initial_rating": 5,
"confidence": 3,
"soundness": 4,
"contribution": 2,
"presentation": 4,
"summary": "The authors propose a new definition for reward hacking, based on the idea that we should only care about hacking of proxy reward functions which are correlated (... | {
"rating": "5;5;6;6;8",
"rating_avg": 6,
"confidence": "3;3;3;4;4",
"confidence_avg": 3.4,
"soundness": "3;4;3;3;4",
"soundness_avg": 3.4,
"contribution": "3;2;3;3;4",
"contribution_avg": 3,
"presentation": "2;4;4;4;4",
"presentation_avg": 3.6
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.481749"
} | {
"id": "TCm0Mw9ZFd",
"metareview": "The reviewers agreed that the paper investigates an important problem of reward hacking, proposes a novel definition, and then develops a new regularization that can effectively prevent reward hacking. However, the reviewers also raised several concerns and questions in their in... | {
"decision": "Accept (Spotlight)"
} |
mtyYWBx2ZF | 2410.19542v2 | Brain-like Functional Organization within Large Language Models | {
"content": "## Abstract\n\nAbstract The human brain has long inspired the pursuit of artificial intelligence (AI). Recently, neuroimaging studies provide compelling evidence of alignment between the computational representation of artificial neural networks (ANNs) and the neural responses of the human brain to exte... | [
{
"id": "HkMwTepQjg",
"initial_rating": 1,
"confidence": 4,
"soundness": 1,
"contribution": 1,
"presentation": 3,
"summary": "This paper utilizes a sparse representation method to extract representative response patterns from multiple LLMs. These are then used to create voxel-wise encodi... | {
"rating": "1;3;5;6",
"rating_avg": 3.75,
"confidence": "4;4;4;2",
"confidence_avg": 3.5,
"soundness": "1;3;2;3",
"soundness_avg": 2.25,
"contribution": "1;1;2;3",
"contribution_avg": 1.75,
"presentation": "3;1;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": "2026-01-14T22:16:03.482695"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
myYKk4Qz3l | 2405.17069v1 | Training-free Editioning of Text-to-Image Models | {
"content": "## Abstract\n\nAbstract Inspired by the software industry’s practice of offering different editions or versions of a product tailored to specific user groups or use cases, we propose a novel task, namely, training-free editioning , for text-to-image models.\nSpecifically, we aim to create variations of ... | [
{
"id": "oBu72UoVWi",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper proposes a new task named image editioning that only allows editing a specific concept without training the text-to-image model. To achieve this, this ... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "5;5;4;4",
"confidence_avg": 4.5,
"soundness": "1;3;3;2",
"soundness_avg": 2.25,
"contribution": "1;2;3;2",
"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": "2026-01-14T22:16:03.483332"
} | {
"id": "aXFTiWoe67",
"metareview": "This paper aims to study a new task which only allows editing a specific concept without training the text-to-image model. The authors propose an image editioning method to provide controllable image generation for specific user groups or uses cases. Experiments are performed to... | {
"decision": "Reject"
} |
myolhJPuRI | 2410.15391v1 | Layout-your-3D: Controllable and Precise 3D Generation with 2D Blueprint | {
"content": "## Abstract\n\nAbstract We present Layout-Your-3D, a framework that allows controllable and compositional 3D generation from text prompts.\nExisting text-to-3D methods often struggle to generate assets with plausible object interactions or require tedious optimization processes.\nTo address these challe... | [
{
"id": "VDPNKirETv",
"initial_rating": 5,
"confidence": 2,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "- They enforce structural control of a 3D scene using a 2D layout. Their optimization process avoids collisions.\n- Their method works by first generating a 2D im... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "3;4;2;3",
"confidence_avg": 3,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;2;2",
"contribution_avg": 2,
"presentation": "2;2;2;3",
"presentation_avg": 2.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.483919"
} | {
"id": "YMD6ayHIJG",
"metareview": "The paper tackles the problem of controllable 3D scene generation. The proposed pipeline starts with 2D layout generation and then progresses through 3D instance reconstruction and arrangement. Their key insight is to decompose the original complex problem into a series of simpl... | {
"decision": "Accept (Poster)"
} |
mzJAupYURK | 2410.18958v1 | Stable Consistency Tuning: Understanding and Improving Consistency Models | {
"content": "## Abstract\n\nAbstract Diffusion models achieve superior generation quality but suffer from slow generation speed due to the iterative nature of denoising. In contrast, consistency models, a new generative family, achieve competitive performance with significantly faster sampling.\nThese models are tra... | [
{
"id": "EqjA4IZi9n",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This paper analyzes the training bottlenecks of consistency models and proposes several improvements for better training of consistency models which are based on ... | {
"rating": "1;3;3;5",
"rating_avg": 3,
"confidence": "4;5;5;4",
"confidence_avg": 4.5,
"soundness": "1;2;2;3",
"soundness_avg": 2,
"contribution": "1;2;2;2",
"contribution_avg": 1.75,
"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": "2026-01-14T22:16:03.484615"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
n0OtGl6VGb | 2407.21018v2 | ThinK: Thinner Key Cache by Query-Driven Pruning | {
"content": "## Abstract\n\nAbstract Large Language Models (LLMs) have revolutionized the field of natural language processing, achieving unprecedented performance across a variety of applications.\nHowever, their increased computational and memory demands present significant challenges, especially when handling lon... | [
{
"id": "cW0lsMLQeT",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The paper presents a KV cache optimization method with channel wise pruning for reducing the inference cost under long context scenario. The method is based on th... | {
"rating": "5;5;5;6;8",
"rating_avg": 5.8,
"confidence": "5;4;4;4;3",
"confidence_avg": 4,
"soundness": "3;2;3;3;3",
"soundness_avg": 2.8,
"contribution": "2;2;2;3;3",
"contribution_avg": 2.4,
"presentation": "3;2;3;3;3",
"presentation_avg": 2.8
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.485263"
} | {
"id": "wrY1OFDcgk",
"metareview": "The paper introduces a novel query-dependent KV cache pruning method to reduce memory usage during LLM inference. By targeting channel sparsity in the Key cache, ThinK achieves over 20% memory reduction and 2.8× peak memory savings with minimal accuracy loss, validated across ex... | {
"decision": "Accept (Spotlight)"
} |
n2NidsYDop | 2410.08633v1 | Transformers Provably Solve Parity Efficiently with Chain of Thought | {
"content": "## Abstract\n\nAbstract This work provides the first theoretical analysis of training transformers to solve complex problems by recursively generating intermediate states, analogous to fine-tuning for chain-of-thought (CoT) reasoning. We consider training a one-layer transformer to solve the fundamental... | [
{
"id": "dAssu7up0z",
"initial_rating": 8,
"confidence": 3,
"soundness": 4,
"contribution": 4,
"presentation": 4,
"summary": "The paper provides a theoretical foundations which explain the benefits of chain-of-though. The authors study a simple setup of k-parity problem with 1-layer Tran... | {
"rating": "6;8;8",
"rating_avg": 7.333333333333333,
"confidence": "4;3;3",
"confidence_avg": 3.3333333333333335,
"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 Oral",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.486014"
} | {
"id": "d6v4detEBq",
"metareview": "The submission studies a simple setup of k-parity problem with 1-layer Transformer and provides a separation results for transformer trained without intermediate supervision and one trained with teacher forcing, thereby showing the importance of chain-of-thought.\n\n+ The paper ... | {
"decision": "Accept (Oral)"
} |
n4HH7g9hxk | 2402.03055v2 | Deep Exploration with PAC-Bayes | {
"content": "## Abstract\n\nAbstract Reinforcement learning for continuous control under sparse rewards is an under-explored problem despite its significance in real life. Many complex skills build on intermediate ones as prerequisites. For instance, a humanoid locomotor has to learn how to stand before it can learn... | [
{
"id": "U8PcvZ4YcS",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This paper studies RL in continuous spaces, and is specifically motivated by sparse-reward problems that require exploration. They take a PAC-Bayesian perspective... | {
"rating": "3;5;5;6",
"rating_avg": 4.75,
"confidence": "3;3;2;3",
"confidence_avg": 2.75,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;3;2;3",
"contribution_avg": 2.5,
"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": "2026-01-14T22:16:03.486988"
} | {
"id": "Ch2Aucfikr",
"metareview": "This paper derives a deep exploration method from the perspective of PAC-Bayes. They treat an ensemble as an empirical estimate of a posterior, and apply posterior sampling during training time for exploration and Bayesian model averaging during evaluation time. They show result... | {
"decision": "Reject"
} |
n4wcdct43X | 2402.05187v2 | Learning mirror maps in policy mirror descent | {
"content": "## Abstract\n\nAbstract Policy Mirror Descent (PMD) is a popular framework in reinforcement learning, serving as a unifying perspective that encompasses numerous algorithms. These algorithms are derived through the selection of a mirror map and enjoy finite-time convergence guarantees. Despite its popul... | [
{
"id": "LsUFQi9Sxw",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper studies the impact of different mirror graphs within the Policy Mirror Descent (PMD) framework in reinforcement learning, traditionally dominated by ne... | {
"rating": "5;5;8",
"rating_avg": 6,
"confidence": "3;3;4",
"confidence_avg": 3.3333333333333335,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "3;3;3",
"contribution_avg": 3,
"presentation": "4;2;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.487742"
} | {
"id": "1x8PtpNh28",
"metareview": "In this paper, the authors argue that the well-studied negative entropy mirror map, which results in natural policy gradient, is often sub-optimal. They empirically show that learning a proper mirror map using extensive search methods, such as evolutionary strategies, \ncan re... | {
"decision": "Accept (Poster)"
} |
n64NYyc6rQ | 2406.05127v3 | Towards Semantic Equivalence of Tokenization in Multimodal LLM | {
"content": "## Abstract\n\nAbstract Multimodal Large Language Models (MLLMs) have demonstrated exceptional capabilities in processing vision-language tasks.\nOne of the crux of MLLMs lies in vision tokenization, which involves efficiently transforming input visual signals into feature representations that are most ... | [
{
"id": "SKorpbD1mX",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This work purpose a novel dynamic Semantic-Equivalent Vision Tokenizer(SeTok)，using adynamic cluster method to group visual features into semantic units, , flexib... | {
"rating": "5;6;6;6;6",
"rating_avg": 5.8,
"confidence": "5;5;4;4;4",
"confidence_avg": 4.4,
"soundness": "3;2;3;2;3",
"soundness_avg": 2.6,
"contribution": "3;3;3;3;3",
"contribution_avg": 3,
"presentation": "2;3;3;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.488713"
} | {
"id": "nbkqflDZEV",
"metareview": "This work purposes a novel dynamic Semantic-Equivalent Vision Tokenizer (SeTok)，using a dynamic cluster method to group visual features into semantic units, flexibly determining the number of tokens based on image complexity. The proposed SETOK-based MLLM (SETOKIM) shows superio... | {
"decision": "Accept (Poster)"
} |
n6To2wAOKL | 2406.05630v2 | Ctrl-V: Higher Fidelity Video Generation with Bounding-Box Controlled Object Motion | {
"content": "## Abstract\n\nAbstract With recent advances in video prediction, controllable video generation has been attracting more attention. Generating high fidelity videos according to simple and flexible conditioning is of particular interest. To this end, we propose a controllable video generation model using... | [
{
"id": "6PzltAsxr0",
"initial_rating": 6,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper presents a controllable video generation pipeline for autonomous driving, where bounding boxes are first generated and then used as conditions for vide... | {
"rating": "3;3;3;6",
"rating_avg": 3.75,
"confidence": "5;4;3;5",
"confidence_avg": 4.25,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "2;2;1;3",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.489515"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
n72WC4a9qQ | 2406.06947v2 | CAAP: Context-Aware Action Planning Prompting to Solve Computer Tasks with Front-End UI Only | {
"content": "## Abstract\n\nAbstract Software robots have long been used in Robotic Process Automation (RPA) to automate mundane and repetitive computer tasks. With the advent of Large Language Models (LLMs) and their advanced reasoning capabilities, these agents are now able to handle more complex or previously uns... | [
{
"id": "QsFwGP6lWu",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces CAAP, which is a novel framework for automating complex computer tasks using LLMs by interacting solely through the front-end UI. Unlike pre... | {
"rating": "3;5;5",
"rating_avg": 4.333333333333333,
"confidence": "3;3;3",
"confidence_avg": 3,
"soundness": "2;3;3",
"soundness_avg": 2.6666666666666665,
"contribution": "1;2;3",
"contribution_avg": 2,
"presentation": "2;3;3",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.490157"
} | {
"id": "iACQYyBAyV",
"metareview": "The paper proposes a modular LLM-based agent for solving GUI-based tasks using screenshots. The work addresses an important problem and demonstrates impressive engineering efforts. The proposed method achieves good performance on two benchmarks. \n\nHowever, reviewers raised sig... | {
"decision": "Reject"
} |
n7n8McETXw | 2410.02167v2 | Training Nonlinear Transformers for Chain-of-Thought Inference: A Theoretical Generalization Analysis | {
"content": "## Abstract\n\nAbstract Chain-of-Thought (CoT) is an efficient prompting method that enables the reasoning ability of large language models by augmenting the query using multiple examples with multiple intermediate steps. Despite the empirical success, the theoretical understanding of how to train a Tra... | [
{
"id": "LiODM3JT8r",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "This work presented a theoretical study on training Transformers to achieve Chain-of-Thought (CoT) reasoning capabilities. The authors explored the training dynam... | {
"rating": "5;6;6;8",
"rating_avg": 6.25,
"confidence": "3;2;3;2",
"confidence_avg": 2.5,
"soundness": "2;3;3;4",
"soundness_avg": 3,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.491322"
} | {
"id": "V4UV9u4JuM",
"metareview": "Summary: The paper investigates how nonlinear Transformers can be trained to perform Chain-of-Thought (CoT) reasoning, providing a theoretical foundation for this capability. It examines the training dynamics required to develop CoT abilities and demonstrates that Transformers t... | {
"decision": "Accept (Poster)"
} |
n7qGCmluZr | 2402.04355v1 | PQMass: Probabilistic Assessment of the Quality of Generative Models using Probability Mass Estimation | {
"content": "## Abstract\n\nAbstract We propose a comprehensive sample-based method for assessing the quality of generative models. The proposed approach enables the\nestimation of the probability that two sets of samples are drawn from the same distribution,\nproviding a statistically rigorous method for assessing ... | [
{
"id": "19KAQsE4Ct",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces PQMass, a likelihood-free approach for assessing the quality of generative models by comparing sample distributions without estimating probab... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "4;4;2;3",
"confidence_avg": 3.25,
"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": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.492265"
} | {
"id": "MbAjgDLEHd",
"metareview": "This paper suggests a method to assess how good a generative model is by a likelihood free probability mass estimation procedure. While I find some aspects heuristic, the reviewers are in agreement that there are good ideas here and their opinions improved after the rebuttal per... | {
"decision": "Accept (Poster)"
} |
n8IzL0Vy4G | 2407.13048v1 | Establishing Knowledge Preference in Language Models | {
"content": "## Abstract\n\nAbstract Language models are known to encode a great amount of factual knowledge through pretraining. However, such knowledge might be insufficient to cater to user requests, requiring the model to integrate external knowledge sources and adhere to user-provided specifications.\nWhen answ... | [
{
"id": "frspdeBrfC",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The paper proposed a new problem setting of knowledge preference, which attempts to unify knowledge editing and RAG. The paper proposed a benchmark for the proble... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "3;4;3;3",
"confidence_avg": 3.25,
"soundness": "2;2;3;3",
"soundness_avg": 2.5,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"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": "2026-01-14T22:16:03.492989"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
n8O0trhost | 2407.01574v1 | cryoSPHERE: Single-Particle HEterogeneous REconstruction from cryo EM | {
"content": "## Abstract\n\nAbstract The three-dimensional structure of a protein plays a key role in determining its function. Methods like AlphaFold have revolutionized protein structure prediction based only on the amino-acid sequence. However, proteins often appear in multiple different conformations, and it is ... | [
{
"id": "syr25MFaXZ",
"initial_rating": 6,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "The authors introduced a new method, cryoSPHERE, that resolves the continuous heterogeneity problem in cryo-EM reconstruction. Similar to some other methods, cryo... | {
"rating": "3;3;8",
"rating_avg": 4.666666666666667,
"confidence": "4;4;4",
"confidence_avg": 4,
"soundness": "3;2;3",
"soundness_avg": 2.6666666666666665,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "2;2;3",
"presentation_avg": 2.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.494081"
} | {
"id": "qYpAhd4BeC",
"metareview": "The paper proposes cryoSPHERE, a method for reconstructing proteins that can occur in different configurations. \n\nThe paper received three reviews, and after the rebuttal all three reviewers are supportive of publication.\n\n* Reviewer RJxo was initially not supportive of publ... | {
"decision": "Accept (Poster)"
} |
n8h1z588eu | 2411.01115v1 | Relax and Merge: A Simple Yet Effective Framework for Solving Fair $k$-Means and $k$-sparse Wasserstein Barycenter Problems | {
"content": "## Abstract\n\nAbstract The fairness of clustering algorithms has gained widespread attention across various areas in machine learning. In this paper, we study fair k 𝑘 k italic_k -means clustering in Euclidean space.\nGiven a dataset comprising several groups, the fairness constraint requires that eac... | [
{
"id": "YKKQGM2W0n",
"initial_rating": 8,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "$ (\\alpha, \\beta) $ -fair k-means clustering is a group-based notion of fairness where the number of points belonging to each protected 'group' (race, gender et... | {
"rating": "6;6;8;8",
"rating_avg": 7,
"confidence": "4;4;5;3",
"confidence_avg": 4,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.494930"
} | {
"id": "hQG1ykRTUS",
"metareview": "The notion of $(\\alpha, \\beta)$ fair clustering was introduced to consider $k$-means and other clustering algorithms with fairness constraints. The paper provides improved approximation algorithms for this problem and also the problem of $k$-sparse Wasserstein Barycenter. The ... | {
"decision": "Accept (Poster)"
} |
nA464tCGR5 | 2410.10174v2 | Balanced Neural ODEs: nonlinear model order reduction and Koopman operator approximations | {
"content": "## Abstract\n\nAbstract Variational Autoencoders (VAEs) are a powerful framework for learning compact latent representations, while Neural ODEs excel in learning transient system dynamics.\nThis work combines the strengths of both to create fast surrogate models with adjustable complexity.\nBy leveragin... | [
{
"id": "cL78uafcuY",
"initial_rating": 8,
"confidence": 3,
"soundness": 4,
"contribution": 3,
"presentation": 2,
"summary": "A method for model order reduction using $\\beta$-VAEs and state space NeuralODEs called **balanced neural ODE** is described and tested on a few examples. The hy... | {
"rating": "3;3;5;8",
"rating_avg": 4.75,
"confidence": "2;3;3;3",
"confidence_avg": 2.75,
"soundness": "3;2;3;4",
"soundness_avg": 3,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "1;2;2;2",
"presentation_avg": 1.75
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.495959"
} | {
"id": "rsQmUOX58L",
"metareview": "The authors introduce a new method called Balanced Neural ODEs (B-NODE) which combine $\\beta$-VAE and neural ODEs for modelling dynamical systems. Two primary applications are considered: state reduction and Koopman operator approximation. Reviewers appreciated the impressive b... | {
"decision": "Accept (Poster)"
} |
nAVejJURqZ | 2410.19702v1 | TimeSuite: Improving MLLMs for Long Video Understanding via Grounded Tuning | {
"content": "## Abstract\n\nAbstract Multimodal Large Language Models (MLLMs) have demonstrated impressive performance in short video understanding. However, understanding long-form videos still remains challenging for MLLMs.\nThis paper proposes TimeSuite , a collection of new designs to adapt the existing short-fo... | [
{
"id": "OSgbkIxwrQ",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The paper introduces TimeSuite, a set of designs to adapt short-form video multimodal large language models (MLLMs) for long video understanding. It includes a ne... | {
"rating": "3;6;6;6;6",
"rating_avg": 5.4,
"confidence": "5;4;5;4;4",
"confidence_avg": 4.4,
"soundness": "2;2;3;3;3",
"soundness_avg": 2.6,
"contribution": "3;2;2;3;3",
"contribution_avg": 2.6,
"presentation": "2;2;2;3;3",
"presentation_avg": 2.4
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.496705"
} | {
"id": "AMao7PPoWe",
"metareview": "After discussion, this submission received 3 positive scores and a negative score. The reviewer who assigned the negative score has no response to the author’s rebuttal. After reading the paper, the review comments and the rebuttal, the AC thinks that the major concerns about te... | {
"decision": "Accept (Poster)"
} |
nAl4bz09Mv | 2409.14307v2 | DilateQuant: Accurate and Efficient Diffusion Quantization via Weight Dilation | {
"content": "## Abstract\n\nAbstract Diffusion models have shown excellent performance on various image generation tasks, but the substantial computational costs and huge memory footprint hinder their low-latency applications in real-world scenarios.\nQuantization is a promising way to compress and accelerate models... | [
{
"id": "Md9wHa2BqE",
"initial_rating": 6,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The authors proposed DilateQuant framework to quantize the diffiusion model. First, they propose Weight Dilation (WD) that maximally dilates the unsaturated in-ch... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "4;4;4;4",
"confidence_avg": 4,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"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": "2026-01-14T22:16:03.497404"
} | {
"id": "qm6PGayfg2",
"metareview": "The paper presents DilateQuant, a quantization framework for diffusion models featuring three key components: Weight Dilation (WD), Temporal Parallel Quantizer (TPQ), and Block-wise Knowledge Distillation (BKD). The overall objective is to improve quantization efficiency while m... | {
"decision": "Reject"
} |
nD5tbHBfut | 2306.02928v3 | LRVS-Fashion: Extending Visual Search with Referring Instructions | {
"content": "## Abstract\n\nAbstract This paper introduces a new challenge for image similarity search in the context of fashion, addressing the inherent ambiguity in this domain stemming from complex images. We present Referred Visual Search (RVS), a task allowing users to define more precisely the desired similari... | [
{
"id": "3mro9k2rxp",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "This paper introduces the Referred Visual Search (RVS) task, addressing the challenges of image similarity search in the fashion domain, particularly in scenarios... | {
"rating": "3;5;5;8",
"rating_avg": 5.25,
"confidence": "5;5;4;4",
"confidence_avg": 4.5,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;3;3;4",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.498183"
} | {
"id": "N6lDa8WVWx",
"metareview": "This paper was reviewed by four experts in the field. The paper received mixed review ratings of 8553. \n\nReviewer DWgA is the most positive reviewer, and they liked the proposed e2e method. The major concerns from other reviewers are:\n1. The positioning of the work, being a s... | {
"decision": "Reject"
} |
nDvgHIBRxQ | 2407.08733v2 | Is Your Model Really A Good Math Reasoner? Evaluating Mathematical Reasoning with Checklist | {
"content": "## Abstract\n\nAbstract Exceptional mathematical reasoning ability is one of the key features that demonstrate the power of large language models (LLMs). How to comprehensively define and evaluate the mathematical abilities of LLMs, and even reflect the user experience in real-world scenarios, has emerg... | [
{
"id": "wNNlsdw6IJ",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 4,
"summary": "This paper presents MathCheck, a benchmark/checklist for testing (M)LLMs' task generalization and reasoning robustness in mathematical problems. It utilizes GSM8k... | {
"rating": "3;6;6;8",
"rating_avg": 5.75,
"confidence": "4;3;4;4",
"confidence_avg": 3.75,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;3;3;3",
"contribution_avg": 2.75,
"presentation": "4;3;3;4",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.499373"
} | {
"id": "J2fQprrx4y",
"metareview": "(a) Scientific Claims:\nThe paper introduces MathCheck, a framework for evaluating mathematical reasoning in LLMs through task generalization and reasoning robustness. Using this framework, they develop MathCheck-GSM and MathCheck-GEO for text and multimodal evaluation respectiv... | {
"decision": "Accept (Poster)"
} |
nEDToD1R8M | 2410.07303v2 | Rectified Diffusion: Straightness Is Not Your Need in Rectified Flow | {
"content": "## Abstract\n\nAbstract Diffusion models have greatly improved visual generation but are hindered by slow generation speed due to the computationally intensive nature of solving generative ODEs. Rectified flow, a widely recognized solution, improves generation speed by straightening the ODE path. Its\nk... | [
{
"id": "BT7NbWZlo1",
"initial_rating": 6,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper proposes Rectified Diffusion that generalizes rectified flow to broader categories of diffusion models, improving image generation efficiency by removi... | {
"rating": "1;6;6;8",
"rating_avg": 5.25,
"confidence": "4;3;5;4",
"confidence_avg": 4,
"soundness": "1;3;2;3",
"soundness_avg": 2.25,
"contribution": "1;3;2;3",
"contribution_avg": 2.25,
"presentation": "2;3;2;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.500149"
} | {
"id": "2qJ7f9n9Xa",
"metareview": "This submission aims to generalise rectification to a wider class of diffusion models via first-order ODEs. Authors propose that precomputed noise-data pairs constitute the essential component (c.f. flow-matching, v-prediction), to enable a learned diffusion ODE to be closer to ... | {
"decision": "Accept (Poster)"
} |
nFVsK3QLgs | 2407.11144v1 | YouTube-SL-25: A Large-Scale, Open-Domain Multilingual Sign Language Parallel Corpus | {
"content": "## Abstract\n\nAbstract Even for better-studied sign languages like American Sign Language (ASL), data is the bottleneck for machine learning research. The situation is worse yet for the many other sign languages used by Deaf/Hard of Hearing communities around the world. In this paper, we present YouTub... | [
{
"id": "7XcCKCtzf1",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This paper introduces YouTube-SL-25, a large-scale, open-domain, multilingual sign language video corpus with seemingly well-aligned captions. It features over 3,... | {
"rating": "5;6;8",
"rating_avg": 6.333333333333333,
"confidence": "4;3;4",
"confidence_avg": 3.6666666666666665,
"soundness": "3;3;3",
"soundness_avg": 3,
"contribution": "2;3;4",
"contribution_avg": 3,
"presentation": "3;3;2",
"presentation_avg": 2.6666666666666665
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.500934"
} | {
"id": "6N9105ctXK",
"metareview": "The article has received evaluations from three reviewers, all of which are positive. The authors have also provided detailed responses to the related issues. Therefore, the article will be accepted.",
"additional_comments": "The article has received evaluations from three rev... | {
"decision": "Accept (Poster)"
} |
nGiGXLnKhl | 2403.02308v2 | Vision-RWKV: Efficient and Scalable Visual Perception with RWKV-Like Architectures | {
"content": "## Abstract\n\nAbstract † † * Equal contribution; 🖂 Corresponding author (wangwenhai362@gmail.com) Transformers have revolutionized computer vision and natural language processing, but their high computational complexity limits their application in high-resolution image processing and long-context an... | [
{
"id": "IPjRrz7Qcr",
"initial_rating": 6,
"confidence": 2,
"soundness": 4,
"contribution": 2,
"presentation": 4,
"summary": "The paper introduces a new network architecture, VISION-RWKV, a vision-adapted version of the RWKV network from the NLP community, which employs an RNN-based line... | {
"rating": "5;6;8",
"rating_avg": 6.333333333333333,
"confidence": "3;2;3",
"confidence_avg": 2.6666666666666665,
"soundness": "3;4;3",
"soundness_avg": 3.3333333333333335,
"contribution": "2;2;3",
"contribution_avg": 2.3333333333333335,
"presentation": "4;4;3",
"presentation_avg": 3.66666666666666... | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Spotlight",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.501754"
} | {
"id": "J8nhc23s2e",
"metareview": "(a) Scientific Claims and Findings\n\nThe paper presents VISION-RWKV, an adaptation of the RWKV network from NLP to vision tasks, incorporating an RNN-based linear attention mechanism. Reviewer QwNu notes that the authors have made necessary modifications to suit visual tasks, w... | {
"decision": "Accept (Spotlight)"
} |
nIEjY4a2Lf | 2407.13622v1 | Misspecified $Q$-Learning with Sparse Linear Function Approximation: Tight Bounds on Approximation Error | {
"content": "## Abstract\n\nAbstract The recent work by Dong and Yang ( 2023 ) showed for misspecified sparse linear bandits, one can obtain an O ( ϵ ) 𝑂 italic-ϵ O\\left(\\epsilon\\right) italic_O ( italic_ϵ ) -optimal policy using a polynomial number of samples when the sparsity is a constant, where ϵ italic-ϵ ... | [
{
"id": "e4R7khNS1D",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The authors found a misspecified sparse linear bandit algorithm based on elimination. They first proved that the traditional approaches such as OLIVE work subopti... | {
"rating": "6;6;6;6",
"rating_avg": 6,
"confidence": "3;3;4;3",
"confidence_avg": 3.25,
"soundness": "3;3;4;3",
"soundness_avg": 3.25,
"contribution": "2;2;2;3",
"contribution_avg": 2.25,
"presentation": "3;2;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.502505"
} | {
"id": "vP1UUPoBW9",
"metareview": "This paper present a elimination-based algorithm and polynomial sample complexity bound for reinforcement learning with sparse linear function approximation. The sample complexity is polynomial in the feature dimension d and planning horizon H when the sparsity level is constant... | {
"decision": "Accept (Poster)"
} |
nKSkM5h2VN | 2410.16805v1 | Test-Time Adversarial Defense with Opposite Adversarial Path and high Attack time cost | {
"content": "## Abstract\n\nAbstract Deep learning models are known to be vulnerable to adversarial attacks by injecting sophisticated designed perturbations to input data. Training-time defenses still exhibit a significant performance gap between natural accuracy and robust accuracy. In this paper, we investigate a... | [
{
"id": "pRYauuLk15",
"initial_rating": 5,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper introduces a test-time adversarial defense method that utilizes diffusion-based purification along Opposite Adversarial Paths to excessively denoise ad... | {
"rating": "5;5;5",
"rating_avg": 5,
"confidence": "4;4;3",
"confidence_avg": 3.6666666666666665,
"soundness": "2;2;3",
"soundness_avg": 2.3333333333333335,
"contribution": "3;3;2",
"contribution_avg": 2.6666666666666665,
"presentation": "2;3;2",
"presentation_avg": 2.3333333333333335
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.503339"
} | {
"id": "Px47WI2Hqs",
"metareview": "This work proposed a test-time adversarial defense method via diffusion-based recovery along opposite adversarial paths. \n\nIt received 3 detailed reviews. The ideas and the effectiveness in the reported experiments are recognized by most reviewers. Meanwhile, there are also se... | {
"decision": "Reject"
} |
nNYA7tcJSE | 2410.05651v1 | ViBiDSampler: Enhancing Video Interpolation Using Bidirectional Diffusion Sampler | {
"content": "## Abstract\n\nAbstract Recent progress in large-scale text-to-video (T2V) and image-to-video (I2V) diffusion models has greatly enhanced video generation, especially in terms of keyframe interpolation.\nHowever, current image-to-video diffusion models, while powerful in generating videos from a single ... | [
{
"id": "hGPC5jAbPL",
"initial_rating": 8,
"confidence": 5,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This work introduces a novel, bidirectional sampling strategy to address off-manifold issues without requiring extensive re-noising or fine-tuning. The proposed m... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "3;4;4;4",
"confidence_avg": 3.75,
"soundness": "3;2;3;3",
"soundness_avg": 2.75,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.504012"
} | {
"id": "ILycs3ZRoo",
"metareview": "Summary: Proposes a frame interpolation method that introduces a novel bi-directional sampling approach, that avoids off-manifold issues, in a training-free manner. It sequentially samples along the forward and backward paths, each conditioning on the start and end frames. Addit... | {
"decision": "Accept (Poster)"
} |
nRHD9fAj10 | 2403.18035v3 | Bidirectional Consistency Models | {
"content": "## Abstract\n\nAbstract Diffusion models (DMs) are capable of generating remarkably high-quality samples by iteratively denoising a random vector, a process that corresponds to moving along the probability flow ordinary differential equation (PF ODE).\nInterestingly, DMs can also invert an input image t... | [
{
"id": "ZeGOjcovuN",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "This submission focuses on consistency models, addressing the challenge of bidirectionality by ensuring these models are invertible, allowing a path from data bac... | {
"rating": "5;5;5;6;6",
"rating_avg": 5.4,
"confidence": "4;5;3;3;3",
"confidence_avg": 3.6,
"soundness": "3;2;2;3;3",
"soundness_avg": 2.6,
"contribution": "3;2;4;2;3",
"contribution_avg": 2.8,
"presentation": "3;2;3;4;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.504712"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
nTZOIlf8YH | 2406.00779v1 | Differentiation of Multi-objective Data-driven Decision Pipeline | {
"content": "## Abstract\n\nAbstract Real-world scenarios frequently involve multi-objective data-driven optimization problems, characterized by unknown problem coefficients and multiple conflicting objectives. Traditional two-stage methods independently apply a machine learning model to estimate problem coefficient... | [
{
"id": "Q6H3enQCjl",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 2,
"summary": "This work designs some loss functions by considering solution space, objective space, and decision quality, named landscape loss, Pareto set loss, and decision lo... | {
"rating": "1;3;3",
"rating_avg": 2.3333333333333335,
"confidence": "3;4;3",
"confidence_avg": 3.3333333333333335,
"soundness": "2;2;2",
"soundness_avg": 2,
"contribution": "2;2;2",
"contribution_avg": 2,
"presentation": "2;2;2",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.505619"
} | {
"id": "z74ibqK9HI",
"metareview": "This paper aims to propose a multiobjective decision-focused approach. In order to better align with the inherent properties of multi-objective optimization problems, this paper proposes a set of novel loss functions. These loss functions are designed to capture the discrepancie... | {
"decision": "Reject"
} |
nWO75tVjfp | 2406.06841v2 | CompassDock: Comprehensive Accurate Assessment Approach for Deep Learning-Based Molecular Docking in Inference and Fine-Tuning | {
"content": "## Abstract\n\nAbstract Datasets used for molecular docking, such as PDBBind, contain technical variability - they are noisy. Although the origins of the noise have been discussed, a comprehensive analysis of the physical, chemical, and bioactivity characteristics of the datasets is still lacking. To ad... | [
{
"id": "pE3VZhv4qO",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The present study essentially offers a docking evaluation metric named Comprehensive Accurate Assessment (Compass), mainly focusing on the Physico-Chemical and Bi... | {
"rating": "1;3;3;5",
"rating_avg": 3,
"confidence": "5;4;4;4",
"confidence_avg": 4.25,
"soundness": "1;3;1;3",
"soundness_avg": 2,
"contribution": "1;2;2;2",
"contribution_avg": 1.75,
"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": "2026-01-14T22:16:03.506277"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
nWT6LxbuGi | 2410.13746v1 | Theory on Score-Mismatched Diffusion Models and Zero-Shot Conditional Samplers | {
"content": "## Abstract\n\nAbstract The denoising diffusion model has recently emerged as a powerful generative technique, capable of transforming noise into meaningful data. While theoretical convergence guarantees for diffusion models are well established when the target distribution aligns with the training dist... | [
{
"id": "2J7DkgMc43",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 2,
"presentation": 2,
"summary": "This paper studies the convergence guarantees of zero-shot conditional diffusion models under certain regularity assumptions on the target distributions and assum... | {
"rating": "3;5;5;5;6",
"rating_avg": 4.8,
"confidence": "3;3;3;2;3",
"confidence_avg": 2.8,
"soundness": "2;2;3;3;3",
"soundness_avg": 2.6,
"contribution": "2;3;2;2;2",
"contribution_avg": 2.2,
"presentation": "1;2;3;2;2",
"presentation_avg": 2
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.507659"
} | {
"id": "DcWB68BBvx",
"metareview": "The reviews are all generally on the positive side, though usually not enthusiastically so. The contributions are generally deemed to be valuable, giving convergence analyses for zero-shot conditional diffusion models under suitable regularity conditions, and introducing a zer... | {
"decision": "Accept (Poster)"
} |
nXV3C8aKxZ | 2410.00907v2 | Addition is All You Need for Energy-efficient Language Models | {
"content": "## Abstract\n\nAbstract Large neural networks spend most computation on floating point tensor multiplications. In this work, we find that a floating point multiplier can be approximated by one integer adder with high precision.\nWe propose the linear-complexity multiplication ( ℒ ℒ \\mathcal{L} caligrap... | [
{
"id": "JUxGYGw3tv",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 3,
"presentation": 2,
"summary": "This paper presents L-Mul, an algorithm that approximates floating point multiplications with integer additions, performed at higher precision. The trade-off is f... | {
"rating": "3;5;5;5",
"rating_avg": 4.5,
"confidence": "4;3;4;4",
"confidence_avg": 3.75,
"soundness": "2;3;2;3",
"soundness_avg": 2.5,
"contribution": "2;3;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": "2026-01-14T22:16:03.509207"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
nYEw2KHVxl | 2410.14957v1 | Offline-to-online Reinforcement Learning for Image-based Grasping with Scarce Demonstrations | {
"content": "## Abstract\n\nAbstract Offline-to-online reinforcement learning (O2O RL) aims to obtain a continually improving policy as it interacts with the environment, while ensuring the initial policy behaviour is satisficing.\nThis satisficing behaviour is necessary for robotic manipulation where random explora... | [
{
"id": "0ixTSAsuik",
"initial_rating": 3,
"confidence": 5,
"soundness": 2,
"contribution": 2,
"presentation": 4,
"summary": "This paper introduces a novel offline-to-online reinforcement learning (O2O RL) approach tailored for robotic grasping with limited image-based demonstrations. Th... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "4;5;3;2",
"confidence_avg": 3.5,
"soundness": "3;2;2;2",
"soundness_avg": 2.25,
"contribution": "1;2;2;3",
"contribution_avg": 2,
"presentation": "3;4;3;2",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.510101"
} | {
"id": "EuRsKkccT3",
"metareview": "This paper presents an offline-to-online reinforcement learning approach for image-based robotic grasping, introducing a simplified Q-learning method that replaces the target network with an NTK-inspired regularization technique. Despite some interesting elements, the reviewers ... | {
"decision": "Reject"
} |
nYjAzwor9R | 2410.21107v1 | Tree-Wasserstein Distance for High Dimensional Data with a Latent Feature Hierarchy | {
"content": "## Abstract\n\nAbstract Finding meaningful distances between high-dimensional data samples is an important scientific task. To this end, we propose a new tree-Wasserstein distance (TWD) for high-dimensional data with two key aspects. First, our TWD is specifically designed for data with a latent feature... | [
{
"id": "EQxOZa6mjs",
"initial_rating": 6,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "The paper presents a new Tree-Wasserstein Distance that is specifically designed for a hierarchical data and presents a novel flow matching method. This method ca... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"confidence": "3;3;3;4",
"confidence_avg": 3.25,
"soundness": "2;3;3;2",
"soundness_avg": 2.5,
"contribution": "3;3;3;3",
"contribution_avg": 3,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.511555"
} | {
"id": "rIQRVYaPVE",
"metareview": "The authors focus on estimating the hierarchical structure (i.e., tree structure) given observations from latent hierarchical features, and leverage the tree-Wasserstein for a fast computation for measures with hierarchical structure on supports. The authors propose to embed fea... | {
"decision": "Accept (Poster)"
} |
nbwDsdfJJd | 2411.05750v1 | On Differentially Private String Distances | {
"content": "## Abstract\n\nGiven a database of bit strings A 1 , … , A m ∈ { 0 , 1 } n subscript 𝐴 1 … subscript 𝐴 𝑚 superscript 0 1 𝑛 A_{1},\\ldots,A_{m}\\in\\{0,1\\}^{n} italic_A start_POSTSUBSCRIPT 1 end_POSTSUBSCRIPT , … , italic_A start_POSTSUBSCRIPT italic_m end_POSTSUBSCRIPT ∈ { 0 , 1 } start_POSTSUPERSC... | [
{
"id": "snnVuEYi38",
"initial_rating": 5,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 3,
"summary": "The paper studies the problem of estimating distances between a given set of strings and the query string. The distance metric used in this paper is hamming dista... | {
"rating": "3;3;5;6",
"rating_avg": 4.25,
"confidence": "3;3;3;4",
"confidence_avg": 3.25,
"soundness": "2;3;2;4",
"soundness_avg": 2.75,
"contribution": "2;1;2;3",
"contribution_avg": 2,
"presentation": "2;2;3;3",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.513019"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
ncCuiD3KJQ | 2408.08862v2 | Visual Agents as Fast and Slow Thinkers | {
"content": "## Abstract\n\nAbstract Achieving human-level intelligence requires refining cognitive distinctions between System 1 and System 2 thinking. While contemporary AI, driven by large language models, demonstrates human-like traits, it falls short of genuine cognition. Transitioning from structured benchmark... | [
{
"id": "vk8fB4vnUg",
"initial_rating": 8,
"confidence": 4,
"soundness": 3,
"contribution": 4,
"presentation": 3,
"summary": "Proposes FaST, a Fast-and-Slow-Thinking mechanism for vision-language models that models both “system 1” (fast) and “system 2” (slow) thinking, dynamically switch... | {
"rating": "3;6;8;8",
"rating_avg": 6.25,
"confidence": "3;3;3;4",
"confidence_avg": 3.25,
"soundness": "3;3;3;3",
"soundness_avg": 3,
"contribution": "2;2;3;4",
"contribution_avg": 2.75,
"presentation": "3;3;3;3",
"presentation_avg": 3
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.514039"
} | {
"id": "fS8QKtEeEO",
"metareview": "This work presented a new reasoning mechanism which can adaptively switch between \"system 1\" and \"system 2\" for multimodal LLMs. The authors proposed a top-down reasoning pipeline for visual agent which starts from the global view of the visual input and gradually drill down... | {
"decision": "Accept (Poster)"
} |
nclyFUZpX9 | 2409.20310v2 | A SSM is Polymerized from Multivariate Time Series | {
"content": "## Abstract\n\nAbstract For multivariate time series (MTS) tasks, previous state space models (SSMs) followed the modeling paradigm of Transformer-based methods. However, none of them explicitly model the complex dependencies of MTS: the Channel Dependency variations with Time (CDT). In view of this, we... | [
{
"id": "Gamr9ikAPG",
"initial_rating": 3,
"confidence": 3,
"soundness": 2,
"contribution": 2,
"presentation": 1,
"summary": "This paper proposes Poly-Mamba, a multivariate time series modeling method that extends the standard SSM to multivariate orthogonal basis projection. The author p... | {
"rating": "3;3;5;5",
"rating_avg": 4,
"confidence": "5;3;3;4",
"confidence_avg": 3.75,
"soundness": "2;2;2;2",
"soundness_avg": 2,
"contribution": "1;2;3;2",
"contribution_avg": 2,
"presentation": "1;1;2;1",
"presentation_avg": 1.25
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Conference Withdrawn Submission",
"venueid": "ICLR.cc/2025/Conference/Withdrawn_Submission",
"processed_at": "2026-01-14T22:16:03.515177"
} | {
"id": "",
"metareview": "",
"additional_comments": ""
} | {
"decision": ""
} |
ngmEcEer8a | 2403.17887v1 | The Unreasonable Ineffectiveness of the Deeper Layers | {
"content": "## Abstract\n\nAbstract We\nempirically study\na simple layer-pruning strategy\nfor\npopular\nfamilies\nof open-weight pretrained LLMs,\nfinding minimal degradation of performance on\ndifferent\nquestion-answering benchmarks until\nafter\na large fraction (up to half) of the layers\nare removed.\nTo pru... | [
{
"id": "8mHNAGIOhD",
"initial_rating": 5,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "This study explores how knowledge is stored in large language models (LLMs) by examining the effects of layer pruning. The research identifies layers that can be ... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "3;3;4;4",
"confidence_avg": 3.5,
"soundness": "2;4;3;3",
"soundness_avg": 3,
"contribution": "3;2;2;3",
"contribution_avg": 2.5,
"presentation": "4;4;3;3",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.516035"
} | {
"id": "V9TUzCj6HK",
"metareview": "The paper demonstrates that deeper LLM layers can be pruned without affecting performance on knowledge QA tasks like MMLU, though this doesn't hold for reasoning tasks like GSM-8k. Using layer similarity metrics, they show up to 50% of layers can be removed while maintaining QA ... | {
"decision": "Accept (Poster)"
} |
niywLsa54R | 2411.02572v1 | ViTally Consistent: Scaling Biological Representation Learning for Cell Microscopy | {
"content": "## Abstract\n\nAbstract Large-scale cell microscopy screens are used in drug discovery and molecular biology research to study the effects of millions of chemical and genetic perturbations on cells.\nTo use these images in downstream analysis, we need models that can map each image into a feature space ... | [
{
"id": "p4Q7zf9sn6",
"initial_rating": 3,
"confidence": 4,
"soundness": 3,
"contribution": 2,
"presentation": 3,
"summary": "The authors presented a new 1.9 billion-parameter ViT-G/8 MAE model, trained on over 8 billion microscopy image crops. The authors demonstrated performance boost ... | {
"rating": "1;3;5;8",
"rating_avg": 4.25,
"confidence": "4;4;5;3",
"confidence_avg": 4,
"soundness": "1;3;2;3",
"soundness_avg": 2.25,
"contribution": "1;2;2;3",
"contribution_avg": 2,
"presentation": "4;3;3;4",
"presentation_avg": 3.5
} | {
"primary_area": "",
"track": "main",
"venue": "Submitted to ICLR 2025",
"venueid": "ICLR.cc/2025/Conference/Rejected_Submission",
"processed_at": "2026-01-14T22:16:03.517105"
} | {
"id": "o5DDEghdih",
"metareview": "The paper presents a new unsupervised vision transformer foundation model of cell microscopy with the goal of using the model for downstream tasks including predicting the cellular phenotype of chemical and genetic perturbations. The main contributions are scaling model size, cu... | {
"decision": "Reject"
} |
njvSBvtiwp | 2405.18213v2 | NeRAF: 3D Scene Infused Neural Radiance and Acoustic Fields | {
"content": "## Abstract\n\nAbstract Sound plays a major role in human perception. Along with vision, it provides essential information for understanding our surroundings. Despite advances in neural implicit representations, learning acoustics that align with visual scenes remains a challenge. We propose NeRAF, a me... | [
{
"id": "dgftISjPHH",
"initial_rating": 5,
"confidence": 4,
"soundness": 2,
"contribution": 3,
"presentation": 3,
"summary": "The paper presents NeRAF, a novel framework for synthesizing both vision and spatial audio in 3D scenes (spatial audio is the main focus of this paper). NeRAF’s k... | {
"rating": "5;5;5;6",
"rating_avg": 5.25,
"confidence": "4;5;4;5",
"confidence_avg": 4.5,
"soundness": "2;2;2;3",
"soundness_avg": 2.25,
"contribution": "2;2;3;3",
"contribution_avg": 2.5,
"presentation": "2;3;3;2",
"presentation_avg": 2.5
} | {
"primary_area": "",
"track": "main",
"venue": "ICLR 2025 Poster",
"venueid": "ICLR.cc/2025/Conference",
"processed_at": "2026-01-14T22:16:03.517973"
} | {
"id": "oYXIykLJeQ",
"metareview": "This paper proposes a method (called NeRAF) jointly learning spatial acoustic and radiance fields. NeRAF synthesizes both novel optical views and spatial audio (room impulse responses, RIR) at new positions by conditioning the acoustic field on 3D scene geometric and appearance ... | {
"decision": "Accept (Poster)"
} |
nlpCeFsSYJ | 2410.10429v1 | DOME: Taming Diffusion Model into High-Fidelity Controllable Occupancy World Model | {
"content": "## Abstract\n\nAbstract We propose DOME , a diffusion-based world model that predicts future occupancy frames based on past occupancy observations.\nThe ability of this world model to capture the evolution of the environment is crucial for planning in autonomous driving.\nCompared to 2D video-based worl... | [
{
"id": "MfbnsPdENg",
"initial_rating": 6,
"confidence": 3,
"soundness": 3,
"contribution": 3,
"presentation": 3,
"summary": "The authors proposed a pipeline to predict future occupancy frames conditioned on historical occupancy observations and trajectory control. The approach entails O... | {
"rating": "5;5;6;6",
"rating_avg": 5.5,
"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": "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": "2026-01-14T22:16:03.518724"
} | {
"id": "QhgwNY8x8O",
"metareview": "This paper introduces DOME, a diffusion-based occupancy world model for predicting future occupancy frames conditioned on historical observations and trajectory control. The approach leverages a spatial-temporal diffusion Transformer for long-duration generation and a trajectory... | {
"decision": "Reject"
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.