claim_id large_stringlengths 20 20 | paper_id large_stringlengths 17 20 | year int32 2.02k 2.03k | atomic_claim large_stringlengths 18 387 | claim_index int32 0 8 | extractor_model large_stringclasses 1
value | prompt_hash large_stringclasses 1
value | run_id large_stringclasses 2
values | created_at large_stringclasses 2
values |
|---|---|---|---|---|---|---|---|---|
clm_a48eb9a0c4f68e49 | 2025.emnlp-main.1038 | 2,025 | DiCoRe is a divergent-convergent reasoning framework that decouples the task of event detection using Dreamer and Grounder. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_8d4f7a39c851b67f | 2025.emnlp-main.1038 | 2,025 | Dreamer encourages divergent reasoning through open-ended event discovery to boost event coverage. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_864157ce23cfd8a6 | 2025.emnlp-main.1038 | 2,025 | Grounder introduces convergent reasoning with finite-state machine guided constrained decoding to align free-form predictions with task-specific instructions. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_022379099d606d8b | 2025.emnlp-main.1038 | 2,025 | LLM-Judge verifies final outputs to ensure high precision. | 3 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_c323111d9655f247 | 2025.emnlp-main.1038 | 2,025 | DiCoRe consistently outperforms prior zero-shot, transfer-learning, and reasoning baselines in event detection. | 4 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_13bf6e49027c7c92 | 2021.emnlp-main.357 | 2,021 | FLODIAL is a dataset of flowchart-grounded task-oriented dialogs for troubleshooting. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_0feb92f9bb01055d | 2021.emnlp-main.357 | 2,021 | FLONET is a neural model for flowchart-grounded task-oriented dialog systems that uses a retrieval-augmented generation architecture. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_0ba20437bc9a1d03 | 2021.emnlp-main.357 | 2,021 | FLONET can perform zero-shot transfer to unseen flowcharts. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_4271cbde358fec81 | 2021.emnlp-main.175 | 2,021 | MSKE-Dialog leverages multiple heterogeneous knowledge sources including commonsense knowledge facts, text knowledge, and infobox knowledge to improve the knowledge coverage. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_805c2f004cdc9970 | 2021.emnlp-main.175 | 2,021 | Multi-Reference Selection avoids topic conflict among context and different knowledge sources by better selecting context and knowledge. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_6d474b7753e50656 | 2021.emnlp-main.175 | 2,021 | Multi-Reference Generation generates informative responses by referring to multiple generation references simultaneously. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_bd9d962106aa998a | 2021.emnlp-main.175 | 2,021 | Multi-source heterogeneous knowledge is used in open-domain knowledge-enhanced dialogue generation for the first time. | 3 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_b2df425fea13adc1 | 2025.emnlp-main.875 | 2,025 | Large language models exhibit high susceptibility of Agreeableness, Extraversion, and Conscientiousness to user influence, while Emotional Stability and Intellect remain relatively stable. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_c171e3fadf702ced | 2025.emnlp-main.875 | 2,025 | Large language models dynamically adjust their conversational style in response to user personas. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_371929bb9b036d44 | 2021.emnlp-main.365 | 2,021 | CSDS provides role-oriented summaries to capture different speakers' viewpoints in Chinese customer service dialogues. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_5a144e0449371ee8 | 2021.emnlp-main.365 | 2,021 | CSDS structures summaries at the topic level to represent the topic flow of Chinese customer service dialogues. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_50a29f60c76da768 | 2021.emnlp-main.365 | 2,021 | Existing summarization methods generate redundant and incoherent summaries when evaluated on CSDS. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_03117b4dbdca8e1c | 2021.emnlp-main.365 | 2,021 | Existing summarization methods have lower performance on role-oriented summaries and topic structures when evaluated on CSDS. | 3 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_55d7f87496360707 | 2021.emnlp-main.572 | 2,021 | CrossFit standardizes seen/unseen task partitions, data access during different learning stages, and evaluation protocols for studying cross-task generalization ability. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_ee24e3cbed963dd7 | 2021.emnlp-main.572 | 2,021 | NLP Few-shot Gym provides diverse few-shot NLP tasks created from open-access NLP datasets and converted to a unified text-to-text format. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_c4dc3b11e2b7dc61 | 2021.emnlp-main.572 | 2,021 | Few-shot learning ability on unseen tasks can be improved by an upstream learning stage using a set of seen tasks. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_bf7205eec0596f70 | 2021.emnlp-main.572 | 2,021 | The selection of upstream learning tasks significantly influences few-shot performance on unseen tasks. | 3 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_3f371a56389b5ff2 | 2021.emnlp-main.219 | 2,021 | Synchronous dual network with cross-type attention separately and interactively considers entity types and relation types for joint entity and relation extraction. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_91e8a5dca29d74cf | 2021.emnlp-main.219 | 2,021 | Synchronous dual network uses two isomorphic bi-directional type-attention LSTM to encode entity type enhanced representations and relation type enhanced representations. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_d267ededb19d81bb | 2021.emnlp-main.219 | 2,021 | Synchronous dual network explicitly models the interdependence between entity types and relation types via cross-type attention. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_beccc45cc8a65edb | 2021.emnlp-main.219 | 2,021 | A new multi-task learning strategy models the interaction between entity type information and relation type information. | 3 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_f985a3d188f22548 | 2021.emnlp-main.219 | 2,021 | Synchronous dual network achieves state-of-the-art performance on the NYT and WebNLG datasets for joint entity and relation extraction. | 4 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_ee87322ef802ca19 | 2022.emnlp-main.154 | 2,022 | SafeText is a benchmark dataset for commonsense physical safety consisting of real-life scenarios with paired safe and physically unsafe advice. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_f52bee3d0303a66b | 2022.emnlp-main.154 | 2,022 | State-of-the-art large language models generate advice that violates commonsense physical safety. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_39464ed5df8257c5 | 2022.emnlp-main.154 | 2,022 | State-of-the-art large language models have difficulty rejecting advice that violates commonsense physical safety. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_6f91ebd201adf835 | 2025.emnlp-main.953 | 2,025 | Debate Speech Evaluation requires LLM judges to assess debate speeches for argument strength, relevance, coherence, organization, style, and tone. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_4b7f4cc1b046f0db | 2025.emnlp-main.953 | 2,025 | Debate Speech Evaluation uses a dataset of meticulously annotated debate speeches. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_eab6a3e8b6b4565d | 2025.emnlp-main.953 | 2,025 | Among state-of-the-art LLMs, larger models approximate individual human judgments in some respects but differ substantially in overall judgment behavior for debate speech evaluation. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_c344ea53ebcd1af4 | 2025.emnlp-main.953 | 2,025 | Frontier LLMs may perform at a human level when generating persuasive, opinionated speeches. | 3 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_351c1d2774890b4b | 2025.emnlp-main.232 | 2,025 | MultiAgentESC does not require training. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_ece26a17dc4d1407 | 2025.emnlp-main.232 | 2,025 | MultiAgentESC emulates the human-like process of providing emotional support through dialogue analysis, strategy deliberation, and response generation. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_9596c74aa339dc3d | 2025.emnlp-main.232 | 2,025 | MultiAgentESC uses a multi-agent system at each stage to iteratively enhance information understanding and reasoning through diverse interactions among expert agents. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_9c9e179120d6e70e | 2025.emnlp-main.232 | 2,025 | MultiAgentESC uses a response-centered approach to handle the one-to-many problem in strategy selection by generating diverse responses with multiple valid strategies and selecting the optimal response through multi-agent collaboration. | 3 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_a34323d2126ff868 | 2025.emnlp-main.232 | 2,025 | MultiAgentESC diversifies support strategy selection. | 4 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_982262d6d89aaaf6 | 2020.emnlp-main.129 | 2,020 | MAVEN is a general-domain event detection dataset containing 4,480 Wikipedia documents, 118,732 event mention instances, and 168 event types. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_385ba2dbf1d3ee48 | 2020.emnlp-main.129 | 2,020 | Existing event detection methods achieve lower performance on MAVEN than on small datasets. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_9bb3363e8e3b6928 | 2020.emnlp-main.619 | 2,020 | STATENet screens for suicidal risk in English tweets by incorporating temporal and emotional context from user history. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_37528e79bf71d15e | 2020.emnlp-main.619 | 2,020 | Emotional and temporal contextual cues from social media history improve suicide risk assessment. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_b34c5fbe815da3b5 | 2021.emnlp-main.583 | 2,021 | Question models that use only the input question text approximate the answer confidence scores of state-of-the-art Question Answering systems. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_02c55a51a0d3343b | 2021.emnlp-main.583 | 2,021 | Transformer-based question models are distilled from Transformer-based answering models. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_87862e88e97818a5 | 2021.emnlp-main.583 | 2,021 | Question models approximate the Precision/Recall curves of target Question Answering systems. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_456f00224377b9a8 | 2021.emnlp-main.583 | 2,021 | Question models used as filters trade off lower computation cost for lower Recall in Question Answering systems. | 3 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_5551e2eab9163659 | 2020.emnlp-main.284 | 2,020 | The Within-Between Relation model recognizes lexical-semantic relations between words. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_478a4aa101da9169 | 2020.emnlp-main.284 | 2,020 | The Within-Between Relation model integrates relational and distributional signals to form a sub-space representation for each lexical-semantic relation. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_55d92980facdef43 | 2021.emnlp-main.9 | 2,021 | SumFC is a fine-grained two-stage framework for assessing factual consistency in abstractive summarization. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_0a1a104336a3c7a1 | 2021.emnlp-main.9 | 2,021 | SumFC selects the top-K most relevant sentences from the document for a summary sentence to assess factual consistency. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_86048470a974046e | 2021.emnlp-main.9 | 2,021 | SumFC performs fine-grained factual consistency reasoning at the sentence level and aggregates consistency scores to produce the final assessment. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_da7bf4b829568abd | 2021.emnlp-main.9 | 2,021 | SumFC uses data synthesis to generate training pairs and contrastive loss to identify subtle cues for factual consistency assessment. | 3 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_5f0b4d4fff2c185f | 2021.emnlp-main.9 | 2,021 | SumFC distinguishes detailed differences in factual consistency assessment better than prior methods. | 4 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_bdd1e3fdb2543e27 | 2025.emnlp-main.553 | 2,025 | State-of-the-art large language models struggle to locate the first error step in student solutions even when given access to the reference solution. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_5b4f1cfd04de2df0 | 2025.emnlp-main.553 | 2,025 | Generating an intermediate corrected student solution that aligns more closely with the original student's solution improves performance in locating the first error step in student solutions. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_02a85a9f765ee8bc | 2025.emnlp-main.463 | 2,025 | ComplexTempQA is a large-scale dataset for complex temporal question answering. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_32e10fd326a87b9d | 2025.emnlp-main.463 | 2,025 | ComplexTempQA covers questions spanning multiple decades. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_eeeb3c00837fce13 | 2025.emnlp-main.463 | 2,025 | A taxonomy categorizes temporal questions as attributes (revolving around events), comparisons (revolving around entities), and counting questions (revolving around time periods). | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_f50a49b0769b4a09 | 2025.emnlp-main.463 | 2,025 | ComplexTempQA questions require reasoning capabilities including across-time comparison, temporal aggregation, and multi-hop reasoning involving temporal event ordering and entity recognition. | 3 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_bffbdb5bcaceb0a5 | 2025.emnlp-main.463 | 2,025 | ComplexTempQA provides detailed metadata for each question, including specific time scopes, to enable comprehensive evaluation of temporal reasoning in large language models. | 4 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_885328fc163aec93 | 2020.emnlp-main.398 | 2,020 | Redundancy in pretrained transformer models has two classes: General Redundancy and Task-specific Redundancy. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_e55ddaae51150238 | 2020.emnlp-main.398 | 2,020 | A large majority of neurons in pretrained transformer models are redundant. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_5a077c4aab2ad6d7 | 2020.emnlp-main.398 | 2,020 | A large majority of redundant neurons in pretrained transformer models can be removed during optimization for a downstream task. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_3096f08d89ff944e | 2020.emnlp-main.398 | 2,020 | A feature-based transfer learning procedure for pretrained transformer models maintains high performance while using a very small fraction of the original neurons. | 3 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_aa5ccfdd29b58b20 | 2023.emnlp-main.663 | 2,023 | The textual encoder facilitates cross-modal conversion in end-to-end speech translation, but noise in speech impedes the consistency between text and speech representations. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_cffce716f89f8687 | 2023.emnlp-main.663 | 2,023 | IMTL bridges the modal gap in end-to-end speech translation by mitigating the difference in length and representation. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_b1c9db23244ad86f | 2023.emnlp-main.663 | 2,023 | With additional data, IMTL achieves state-of-the-art results on the MuST-C end-to-end English-to-Spanish speech translation task while requiring less training time than prior methods. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_f3d8870b9fe33c82 | 2025.emnlp-main.159 | 2,025 | Date fragmentation ratio measures how faithfully tokenisers preserve multi-digit date components. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_66f74332cc4a86ec | 2025.emnlp-main.159 | 2,025 | DateAugBench evaluates temporal reasoning on context-based date resolution, format-invariance puzzles, and date arithmetic across historical, contemporary, and future time periods. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_e194549c6fdaf6d1 | 2025.emnlp-main.159 | 2,025 | Large language models use an emergent date-abstraction mechanism to stitch together fragments of month, day, and year components for temporal reasoning. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_7c5a7c9ec08de122 | 2025.emnlp-main.159 | 2,025 | Excessive date fragmentation correlates with accuracy drops on uncommon dates such as historical and futuristic dates. | 3 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_0a98a4d43b54ce5e | 2025.emnlp-main.159 | 2,025 | Larger models heal date fragments faster via the emergent date abstraction mechanism. | 4 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_a64e462099590da6 | 2025.emnlp-main.159 | 2,025 | Large language models follow a reasoning path to assemble date fragments that differs from human interpretation (year → month → day). | 5 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_b85cf5c255b4280c | 2024.emnlp-main.264 | 2,024 | PsyGUARD detects suicide ideation and assesses suicide risk in psychological counseling. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_edd59d4b32ce50c9 | 2024.emnlp-main.264 | 2,024 | Foundational theories form the basis of a detailed taxonomy for detecting suicide ideation. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_7396cd17101aff8b | 2024.emnlp-main.264 | 2,024 | PsySUICIDE is a large-scale, high-quality dataset for suicide detection. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_d13d97b215b5b942 | 2025.emnlp-main.16 | 2,025 | CompKBQA optimizes the fine-tuning process of large language models for generating logical forms by enabling progressive learning of skeleton generation, topic entity generation, and relevant relations generation. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_c77476c9e07461b5 | 2025.emnlp-main.16 | 2,025 | R3 retrieves and incorporates knowledge base information into the process of logical form generation. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_e17926f2cf23fc43 | 2025.emnlp-main.16 | 2,025 | CompKBQA achieves state-of-the-art performance on the WebQSP and CWQ knowledge base question answering benchmarks. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_adaf5baabe3eb664 | 2025.emnlp-main.1464 | 2,025 | CausalTalk is an annotated dataset of Reddit posts about public health for causal language understanding. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_6aebcba56384d502 | 2025.emnlp-main.1464 | 2,025 | CausalTalk supports benchmarking of both discriminative and generative models for causal language understanding. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_0d2dafd8e99b691b | 2025.emnlp-main.1464 | 2,025 | CausalTalk annotations combine domain expert labels with GPT-4o-generated labels that are human-verified. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_9f799ba7202ee51c | 2025.emnlp-main.1464 | 2,025 | CausalTalk bridges fine-grained causal detection and gist-based reasoning in informal text. | 3 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_4d9a0c69ec6f40d5 | 2023.emnlp-main.607 | 2,023 | CRoW is a manually-curated, multi-task benchmark that evaluates the ability of models to apply commonsense reasoning in six real-world NLP tasks. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_b67b4dc5da109f00 | 2023.emnlp-main.607 | 2,023 | CRoW is constructed using a multi-stage data collection pipeline that rewrites examples from existing datasets using commonsense-violating perturbations. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_51306847fbcdb691 | 2023.emnlp-main.607 | 2,023 | NLP systems exhibit a significant performance gap compared to humans on CRoW. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_dfce960e2eb6f21f | 2024.emnlp-main.926 | 2,024 | Segment+ enables language models with limited context windows to process long texts. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_124a234f29c6775c | 2024.emnlp-main.926 | 2,024 | Segment+ manages information flow for long text processing using structured notes. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_4be3fcbb6279e9d3 | 2024.emnlp-main.926 | 2,024 | Segment+ manages information flow for long text processing using a filtering module. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_39a0b2dde11b4bc8 | 2024.emnlp-main.926 | 2,024 | Segment+ produces a controllable system for long text processing. | 3 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_215b303c450e6d67 | 2024.emnlp-main.926 | 2,024 | Segment+ produces an interpretable system for long text processing. | 4 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_77468c1a16d94e9c | 2020.emnlp-main.360 | 2,020 | Multi-Word Expressions are translated by cross-lingual word embeddings when pre-tokenized as single tokens. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_ef224ae1dd98b800 | 2020.emnlp-main.360 | 2,020 | Bilingual word lists for evaluating Multi-Word Expression translation contain Multi-Word Expressions that correspond to single words in another language. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_8e646bf8738a936c | 2020.emnlp-main.360 | 2,020 | Pre-tokenization of Multi-Word Expressions as single tokens outperforms averaging the embeddings of individual tokens for Multi-Word Expression translation. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_d6ad5e85b74df6d2 | 2020.emnlp-main.360 | 2,020 | Pre-tokenization of Multi-Word Expressions as single tokens does not degrade single-word translation performance. | 3 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_8859350ebf3a75d8 | 2024.emnlp-main.351 | 2,024 | SDS-CLIP is a lightweight and sample-efficient distillation method that enhances CLIP's compositional visio-linguistic reasoning. | 0 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_4b48549cbabf12f9 | 2024.emnlp-main.351 | 2,024 | SDS-CLIP fine-tunes CLIP using a distillation objective from large text-to-image generative models. | 1 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_097ca1db6666ca30 | 2024.emnlp-main.351 | 2,024 | SDS-CLIP improves visio-linguistic performance of CLIP models on the Winoground benchmark. | 2 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
clm_7a2eda83d9c4e620 | 2024.emnlp-main.351 | 2,024 | SDS-CLIP improves visio-linguistic performance of CLIP models on the ARO dataset. | 3 | qwen/qwen3-235b-a22b-thinking-2507 | f9a9b434c72c | extract_20260524T085601Z | 2026-05-24T08:56:02.723585+00:00 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.