{"noiseShare":0.4931,"topics":[{"claims":[1382],"deltaPp":null,"id":0,"keywords":["visual","multimodal","vision-language","video","image"],"label":"Vision-Language Model Evaluation","labelFull":"Evaluation and Robustness of Vision-Language Models","papers":556,"papersByYear":[556],"rel":null,"reviewed":true,"share":[15.3252],"size":1382,"slope":null},{"claims":[1253],"deltaPp":null,"id":1,"keywords":["agents","reinforcement learning","policy","reward","skill"],"label":"LLM Agent Skill Learning","labelFull":"Skill Learning and Management for LLM Agents","papers":647,"papersByYear":[647],"rel":null,"reviewed":true,"share":[17.8335],"size":1253,"slope":null},{"claims":[842],"deltaPp":null,"id":2,"keywords":["reasoning","mathematical","chain-of-thought","traces","benchmarks"],"label":"Chain-of-Thought Reasoning","labelFull":"Chain-of-Thought Reasoning, Scaling, Efficiency for LLMs","papers":480,"papersByYear":[480],"rel":null,"reviewed":true,"share":[13.2304],"size":842,"slope":null},{"claims":[396],"deltaPp":null,"id":3,"keywords":["retrieval","query","reranking","documents","relevance"],"label":"Retrieval and Reranking","labelFull":"Retrieval and Reranking for Multi-Stage IR Pipelines","papers":224,"papersByYear":[224],"rel":null,"reviewed":true,"share":[6.1742],"size":396,"slope":null},{"claims":[388],"deltaPp":null,"id":4,"keywords":["speech","audio","acoustic","translation","recognition"],"label":"Speech LLMs","labelFull":"Speech Large Language Models and Benchmarks","papers":133,"papersByYear":[133],"rel":null,"reviewed":true,"share":[3.6659],"size":388,"slope":null},{"claims":[335],"deltaPp":null,"id":5,"keywords":["languages","english","multilingual","cross-lingual","alignment"],"label":"Multilingual LLM Benchmarking","labelFull":"Evaluation of Large Language Models on Multilingual Benchmarks","papers":228,"papersByYear":[228],"rel":null,"reviewed":true,"share":[6.2845],"size":335,"slope":null},{"claims":[323],"deltaPp":null,"id":6,"keywords":["kv","cache","memory","benchmarks","reduces"],"label":"LLM Inference Efficiency","labelFull":"LLM Inference Efficiency: Memory, Compute, Speed","papers":243,"papersByYear":[243],"rel":null,"reviewed":true,"share":[6.6979],"size":323,"slope":null},{"claims":[298],"deltaPp":null,"id":7,"keywords":["dialogue","conversational","multi-turn","user","interactions"],"label":"Multi-Turn Dialogue Evaluation","labelFull":"Benchmarks and Metrics for Multi-Turn Dialogue Evaluation","papers":165,"papersByYear":[165],"rel":null,"reviewed":true,"share":[4.548],"size":298,"slope":null},{"claims":[288],"deltaPp":null,"id":8,"keywords":["attack","success","jailbreak","rate","defenses"],"label":"LLM Jailbreak Attacks","labelFull":"Jailbreak Attacks and Defenses for Large Language Models","papers":146,"papersByYear":[146],"rel":null,"reviewed":true,"share":[4.0243],"size":288,"slope":null},{"claims":[199],"deltaPp":null,"id":9,"keywords":["token","attention","output","usage","activations"],"label":"Token Pruning","labelFull":"Budget-Aware Token Pruning for Efficient LLM Inference","papers":149,"papersByYear":[149],"rel":null,"reviewed":true,"share":[4.1069],"size":199,"slope":null},{"claims":[182],"deltaPp":null,"id":10,"keywords":["question","answering","factual","pairs","multi-hop"],"label":"Question Answering Benchmarks","labelFull":"Question Answering Benchmarks for Multi-hop and Long-Form","papers":134,"papersByYear":[134],"rel":null,"reviewed":true,"share":[3.6935],"size":182,"slope":null},{"claims":[171],"deltaPp":null,"id":11,"keywords":["code","test","correctness","verification","functional"],"label":"LLM Code Verification","labelFull":"Formal Verification and Functional Correctness of LLM-Generated Code","papers":121,"papersByYear":[121],"rel":null,"reviewed":true,"share":[3.3352],"size":171,"slope":null},{"claims":[143],"deltaPp":null,"id":12,"keywords":["cultural","moral","demographic","cues","alignment"],"label":"Cultural Bias in LLMs","labelFull":"Cultural and Demographic Bias Evaluation in LLMs","papers":90,"papersByYear":[90],"rel":null,"reviewed":true,"share":[2.4807],"size":143,"slope":null},{"claims":[138],"deltaPp":null,"id":13,"keywords":["translation","machine","quality","languages","estimation"],"label":"MT Quality Estimation","labelFull":"MT Quality Estimation and Preference Optimization","papers":58,"papersByYear":[58],"rel":null,"reviewed":true,"share":[1.5987],"size":138,"slope":null},{"claims":[117],"deltaPp":null,"id":14,"keywords":["hallucination","lvlms","detection","reduces","mitigation"],"label":"LLM Hallucination","labelFull":"Hallucination Detection, Mitigation, and Evaluation for LLMs","papers":65,"papersByYear":[65],"rel":null,"reviewed":true,"share":[1.7916],"size":117,"slope":null},{"claims":[114],"deltaPp":null,"id":15,"keywords":["distillation","on-policy","student","teacher","knowledge"],"label":"On-Policy Distillation","labelFull":"On-Policy Knowledge Distillation for Large Language Models","papers":77,"papersByYear":[77],"rel":null,"reviewed":true,"share":[2.1224],"size":114,"slope":null},{"claims":[107],"deltaPp":null,"id":16,"keywords":["clinical","patient","medical","records","prediction"],"label":"Clinical LLM Agent Benchmarks","labelFull":"Benchmarks for Clinical LLM Agents","papers":50,"papersByYear":[50],"rel":null,"reviewed":true,"share":[1.3782],"size":107,"slope":null},{"claims":[106],"deltaPp":null,"id":17,"keywords":["judge","human","judgments","agreement","llm"],"label":"LLM-as-Judge Evaluation","labelFull":"LLM-as-Judge Evaluation: Benchmarks and Human Alignment","papers":69,"papersByYear":[69],"rel":null,"reviewed":true,"share":[1.9019],"size":106,"slope":null},{"claims":[101],"deltaPp":null,"id":18,"keywords":["diffusion","masked","decoding","autoregressive","denoising"],"label":"Diffusion Language Model Decoding","labelFull":"Decoding Strategies and Architectures for Diffusion Language Models","papers":54,"papersByYear":[54],"rel":null,"reviewed":true,"share":[1.4884],"size":101,"slope":null},{"claims":[91],"deltaPp":null,"id":19,"keywords":["neurons","activation","steering","layers","representation"],"label":"Activation Steering in Language Models","labelFull":"Activation Steering and Neuron Selection for Controlling LLM Representations","papers":72,"papersByYear":[72],"rel":null,"reviewed":true,"share":[1.9846],"size":91,"slope":null},{"claims":[87],"deltaPp":null,"id":20,"keywords":["f1","baseline","state-of-the-art","supervised","outperforms"],"label":"Efficient Model Adaptation","labelFull":"Efficient Model Adaptation and Training Methods","papers":80,"papersByYear":[80],"rel":null,"reviewed":true,"share":[2.2051],"size":87,"slope":null},{"claims":[86],"deltaPp":null,"id":21,"keywords":["text","ai-generated","sentiment","stylometric","authorship"],"label":"Stylometric AI-Generated Text Detection","labelFull":"Stylometric Detection and Attribution of AI-Generated Text","papers":42,"papersByYear":[42],"rel":null,"reviewed":true,"share":[1.1577],"size":86,"slope":null},{"claims":[84],"deltaPp":null,"id":22,"keywords":["narrative","story","character","events","human"],"label":"Long-Form Narrative Generation","labelFull":"Coherent Long-Form Narrative Generation and Evaluation","papers":39,"papersByYear":[39],"rel":null,"reviewed":true,"share":[1.075],"size":84,"slope":null},{"claims":[78],"deltaPp":null,"id":23,"keywords":["retrieval-augmented generation","multimodal retrieval-augmented","rag","pipelines","systems"],"label":"Robust Retrieval-Augmented Generation","labelFull":"Robustness and Safety for Retrieval-Augmented Generation","papers":58,"papersByYear":[58],"rel":null,"reviewed":true,"share":[1.5987],"size":78,"slope":null},{"claims":[72],"deltaPp":null,"id":24,"keywords":["memory","episodic","retrieval","recall","user"],"label":"Episodic Memory for LLMs","labelFull":"Episodic Memory Retrieval and Management for LLMs","papers":49,"papersByYear":[49],"rel":null,"reviewed":true,"share":[1.3506],"size":72,"slope":null},{"claims":[71],"deltaPp":null,"id":25,"keywords":["preference","optimization","value","direct","subjective"],"label":"Direct Preference Optimization","labelFull":"Direct Preference Optimization for LLM Alignment","papers":50,"papersByYear":[50],"rel":null,"reviewed":true,"share":[1.3782],"size":71,"slope":null},{"claims":[69],"deltaPp":null,"id":26,"keywords":["epistemic","cultural","theory","moral","fairness"],"label":"Fairness in LLMs","labelFull":"Fairness, Bias, and Moral Reasoning in Large Language Models","papers":43,"papersByYear":[43],"rel":null,"reviewed":true,"share":[1.1852],"size":69,"slope":null},{"claims":[64],"deltaPp":null,"id":27,"keywords":["entity","knowledge","graph","extraction","relation"],"label":"Knowledge Graph Construction","labelFull":"Entity Linking and Relation Extraction for Knowledge Graphs","papers":39,"papersByYear":[39],"rel":null,"reviewed":true,"share":[1.075],"size":64,"slope":null},{"claims":[64],"deltaPp":null,"id":28,"keywords":["review","evidence","peer","llm-generated","research"],"label":"Evidence-Based Review Scoring","labelFull":"Evidence-Based Review Scoring and Claim Verification","papers":38,"papersByYear":[38],"rel":null,"reviewed":true,"share":[1.0474],"size":64,"slope":null},{"claims":[63],"deltaPp":null,"id":29,"keywords":["unlearning","forgetting","knowledge","continual","catastrophic"],"label":"Continual Unlearning","labelFull":"Continual Unlearning and Forgetting in LLMs","papers":45,"papersByYear":[45],"rel":null,"reviewed":true,"share":[1.2404],"size":63,"slope":null},{"claims":[63],"deltaPp":null,"id":30,"keywords":["reinforcement","learning","reasoning","rewards","verifiable"],"label":"Reinforcement Learning for Reasoning","labelFull":"Reinforcement Learning for Reasoning in Large Language Models","papers":59,"papersByYear":[59],"rel":null,"reviewed":true,"share":[1.6262],"size":63,"slope":null},{"claims":[59],"deltaPp":null,"id":31,"keywords":["belief","simulation","user","behavioral","student"],"label":"Belief-Aware User Simulation","labelFull":"Belief-Aware User Simulation for LLM Agents","papers":34,"papersByYear":[34],"rel":null,"reviewed":true,"share":[0.9372],"size":59,"slope":null},{"claims":[57],"deltaPp":null,"id":32,"keywords":["personalization","user","preferences","style","persona"],"label":"LLM Personalization","labelFull":"Personalization Methods for Large Language Models","papers":33,"papersByYear":[33],"rel":null,"reviewed":true,"share":[0.9096],"size":57,"slope":null},{"claims":[55],"deltaPp":null,"id":33,"keywords":["optimization","evolutionary","formulation","constraints","modeling"],"label":"LLM-Driven Optimization Formulation","labelFull":"LLM-Driven Optimization Formulation for Combinatorial Tasks","papers":40,"papersByYear":[40],"rel":null,"reviewed":true,"share":[1.1025],"size":55,"slope":null},{"claims":[54],"deltaPp":null,"id":34,"keywords":["calibration","fusion","error","expected","vlms"],"label":"Model Calibration & Fusion","labelFull":"Calibration and Fusion for Large Language and Vision Models","papers":41,"papersByYear":[41],"rel":null,"reviewed":true,"share":[1.1301],"size":54,"slope":null},{"claims":[53],"deltaPp":null,"id":35,"keywords":["molecular","chemical","text representations","foundation","structural"],"label":"Chemistry Foundation Models","labelFull":"Foundation Models for Chemistry and Molecular Tasks","papers":24,"papersByYear":[24],"rel":null,"reviewed":true,"share":[0.6615],"size":53,"slope":null},{"claims":[53],"deltaPp":null,"id":36,"keywords":["citation","scientific","research","search","recommendation"],"label":"Citation Generation and Evaluation for LLMs","labelFull":"Frameworks and Benchmarks for Sub-sentence Citation Generation and Assessment","papers":31,"papersByYear":[31],"rel":null,"reviewed":true,"share":[0.8545],"size":53,"slope":null},{"claims":[51],"deltaPp":null,"id":37,"keywords":["legal","judgment","profiling","investigation","claims"],"label":"Legal Multi-Agent Debate","labelFull":"Multi-Agent Debate Frameworks for Legal Reasoning and Evaluation","papers":37,"papersByYear":[37],"rel":null,"reviewed":true,"share":[1.0198],"size":51,"slope":null},{"claims":[49],"deltaPp":null,"id":38,"keywords":["emotion","affective","posts","multimodal social","social media"],"label":"Emotion-Aware LLMs","labelFull":"Emotion-Aware LLMs for Multimodal Social Media","papers":19,"papersByYear":[19],"rel":null,"reviewed":true,"share":[0.5237],"size":49,"slope":null},{"claims":[48],"deltaPp":null,"id":39,"keywords":["claim","verification","atomic","evidence","nlp"],"label":"Scientific Claim Verification","labelFull":"Scientific Claim Verification and Evidence Retrieval","papers":37,"papersByYear":[37],"rel":null,"reviewed":true,"share":[1.0198],"size":48,"slope":null},{"claims":[47],"deltaPp":null,"id":40,"keywords":["arabic","modern","sentence","parsing","speakers"],"label":"Arabic LLM Benchmarks","labelFull":"Arabic-Centric LLM Benchmarks and Evaluation","papers":19,"papersByYear":[19],"rel":null,"reviewed":true,"share":[0.5237],"size":47,"slope":null},{"claims":[46],"deltaPp":null,"id":41,"keywords":["graph","text-attributed","node","quantized","representation learning"],"label":"Graph Foundation Models","labelFull":"Graph Foundation Models for Text-Attributed Graphs","papers":25,"papersByYear":[25],"rel":null,"reviewed":true,"share":[0.6891],"size":46,"slope":null},{"claims":[46],"deltaPp":null,"id":42,"keywords":["ablation","validate","robustness","finetuning","effectiveness"],"label":"LLM Ablation Studies","labelFull":"Ablation Studies for LLM Fine-Tuning, Safety, and Robustness","papers":41,"papersByYear":[41],"rel":null,"reviewed":true,"share":[1.1301],"size":46,"slope":null},{"claims":[45],"deltaPp":null,"id":43,"keywords":["sql","queries","tables","relational","procedural"],"label":"Execution-Guided Text-to-SQL","labelFull":"Execution-Guided Text-to-SQL: Robustness and Cross-Dialect","papers":25,"papersByYear":[25],"rel":null,"reviewed":true,"share":[0.6891],"size":45,"slope":null},{"claims":[44],"deltaPp":null,"id":44,"keywords":["qa","multi-hop","benchmarks","open-domain","outperforms"],"label":"Retrieval-Augmented Multi-hop QA","labelFull":"Retrieval-Augmented Generation for Multi-hop QA","papers":41,"papersByYear":[41],"rel":null,"reviewed":true,"share":[1.1301],"size":44,"slope":null},{"claims":[43],"deltaPp":null,"id":45,"keywords":["alignment","sequence","establishes","search","foundation"],"label":"LLM Alignment","labelFull":"LLM Alignment: Methods, Data, and Evaluation","papers":34,"papersByYear":[34],"rel":null,"reviewed":true,"share":[0.9372],"size":43,"slope":null},{"claims":[43],"deltaPp":null,"id":46,"keywords":["decoding","speculative","draft","autoregressive","acceptance"],"label":"Speculative Decoding","labelFull":"Speculative Decoding for Efficient LLM Inference","papers":37,"papersByYear":[37],"rel":null,"reviewed":true,"share":[1.0198],"size":43,"slope":null},{"claims":[42],"deltaPp":null,"id":47,"keywords":["privacy","differential","leakage","private","trade-off"],"label":"Differentially Private LLMs","labelFull":"Differential Privacy for Large Language Models","papers":20,"papersByYear":[20],"rel":null,"reviewed":true,"share":[0.5513],"size":42,"slope":null},{"claims":[41],"deltaPp":null,"id":48,"keywords":["annotation","human","agreement","evaluation","nlp"],"label":"Inter-Annotator Agreement","labelFull":"Modeling and Evaluating Inter-Annotator Agreement in NLP","papers":30,"papersByYear":[30],"rel":null,"reviewed":true,"share":[0.8269],"size":41,"slope":null},{"claims":[41],"deltaPp":null,"id":49,"keywords":["perturbations","forecasting","representative","outputs","robust"],"label":"LLM Robustness Benchmarks","labelFull":"Comprehensive Benchmarks for LLM Robustness and Forecasting","papers":37,"papersByYear":[37],"rel":null,"reviewed":true,"share":[1.0198],"size":41,"slope":null},{"claims":[39],"deltaPp":null,"id":50,"keywords":["debate","multi-agent","negotiation","role","diversity"],"label":"Multi-Agent Debate","labelFull":"Multi-Agent Debate: Role Prompting, Diversity, and Alignment","papers":24,"papersByYear":[24],"rel":null,"reviewed":true,"share":[0.6615],"size":39,"slope":null},{"claims":[38],"deltaPp":null,"id":51,"keywords":["summarization","dialectal","fact","case","conversation"],"label":"Long-Form Summarization","labelFull":"Long-Form Summarization: Multimodal & Faithfulness","papers":24,"papersByYear":[24],"rel":null,"reviewed":true,"share":[0.6615],"size":38,"slope":null},{"claims":[38],"deltaPp":null,"id":52,"keywords":["watermarking","linguistic","context-aware","threshold","survey"],"label":"Robust Multilingual LLM Watermarking","labelFull":"Robust Multilingual Watermarking for Large Language Models","papers":15,"papersByYear":[15],"rel":null,"reviewed":true,"share":[0.4135],"size":38,"slope":null},{"claims":[37],"deltaPp":null,"id":53,"keywords":["theorem","formal","proof","physics","rewriting"],"label":"LLM Formal Theorem Proving","labelFull":"LLM-Based Formal Theorem Proving and Autoformalization","papers":24,"papersByYear":[24],"rel":null,"reviewed":true,"share":[0.6615],"size":37,"slope":null},{"claims":[36],"deltaPp":null,"id":54,"keywords":["attention","heads","layers","values","transformer"],"label":"Efficient Hybrid Attention","labelFull":"Efficient Hybrid Attention for Long-Context Transformers","papers":26,"papersByYear":[26],"rel":null,"reviewed":true,"share":[0.7166],"size":36,"slope":null},{"claims":[34],"deltaPp":null,"id":55,"keywords":["retrieval","multimodal","heads","image","fusion"],"label":"Multimodal Retrieval Benchmarks","labelFull":"Benchmarking and Evaluation of Multimodal Retrieval Models","papers":21,"papersByYear":[21],"rel":null,"reviewed":true,"share":[0.5788],"size":34,"slope":null},{"claims":[33],"deltaPp":null,"id":56,"keywords":["counseling","psychological","sessions","professional","synthetic"],"label":"Therapeutic Dialogue Generation","labelFull":"LLM-Based Therapeutic Dialogue Generation and Evaluation","papers":19,"papersByYear":[19],"rel":null,"reviewed":true,"share":[0.5237],"size":33,"slope":null},{"claims":[33],"deltaPp":null,"id":57,"keywords":["reading","surprisal","proficiency","duration","predictive"],"label":"Surprisal Theory for Reading","labelFull":"Surprisal Theory for Modeling Human Reading","papers":11,"papersByYear":[11],"rel":null,"reviewed":true,"share":[0.3032],"size":33,"slope":null},{"claims":[33],"deltaPp":null,"id":58,"keywords":["tokenizer","vocabulary","languages","dialectal","boundaries"],"label":"Multilingual Tokenization","labelFull":"Multilingual Tokenizer Design, Evaluation, and Alignment","papers":19,"papersByYear":[19],"rel":null,"reviewed":true,"share":[0.5237],"size":33,"slope":null},{"claims":[33],"deltaPp":null,"id":59,"keywords":["sparse","autoencoder","merging","features","hierarchical"],"label":"Hierarchical Sparse Autoencoder Merging","labelFull":"Hierarchical Wasserstein Merging of Sparse Autoencoder Features","papers":25,"papersByYear":[25],"rel":null,"reviewed":true,"share":[0.6891],"size":33,"slope":null},{"claims":[33],"deltaPp":null,"id":60,"keywords":["quantization","block","error","post-training","scaling"],"label":"LLM Quantization","labelFull":"Low-bit Quantization Techniques for Large Language Models","papers":22,"papersByYear":[22],"rel":null,"reviewed":true,"share":[0.6064],"size":33,"slope":null},{"claims":[33],"deltaPp":null,"id":61,"keywords":["low-rank","gradients","lora","subspace","orthogonal"],"label":"Subspace Low-Rank Adaptation","labelFull":"Subspace Low-Rank Adaptation for Efficient Fine-Tuning","papers":27,"papersByYear":[27],"rel":null,"reviewed":true,"share":[0.7442],"size":33,"slope":null},{"claims":[31],"deltaPp":null,"id":62,"keywords":["abstention","choice","inflation","social","frontier"],"label":"Conformal Social Choice","labelFull":"Conformal Social Choice for Calibrated LLM Abstention","papers":22,"papersByYear":[22],"rel":null,"reviewed":true,"share":[0.6064],"size":31,"slope":null},{"claims":[30],"deltaPp":null,"id":63,"keywords":["reinforcement learning","learning framework","reward","feedback","trains"],"label":"LLM Reinforcement Learning","labelFull":"Reinforcement Learning for LLM Alignment and Self-Improvement","papers":29,"papersByYear":[29],"rel":null,"reviewed":true,"share":[0.7993],"size":30,"slope":null},{"claims":[28],"deltaPp":null,"id":64,"keywords":["report","medical","radiology","clinical","insights"],"label":"Multimodal Medical Report Generation","labelFull":"Multimodal Medical Report Generation and Benchmarking","papers":16,"papersByYear":[16],"rel":null,"reviewed":true,"share":[0.441],"size":28,"slope":null},{"claims":[27],"deltaPp":null,"id":65,"keywords":["topic","mechanistic","coherence","clustering","neural"],"label":"Mechanistic Topic Modeling","labelFull":"Mechanistic Topic Modeling with Controllable Generation and Evaluation","papers":17,"papersByYear":[17],"rel":null,"reviewed":true,"share":[0.4686],"size":27,"slope":null},{"claims":[27],"deltaPp":null,"id":66,"keywords":["teacher","student","dense","distillation","token-level"],"label":"Dense Token-Level Distillation","labelFull":"Dense Token-Level Distillation for RL Fine-Tuning of LLMs","papers":20,"papersByYear":[20],"rel":null,"reviewed":true,"share":[0.5513],"size":27,"slope":null},{"claims":[27],"deltaPp":null,"id":67,"keywords":["perplexity","sparsity","downstream","state-of-the-art","parameters"],"label":"Sparse Low-Bit LLM Compression","labelFull":"Sparse and Low-Bit Compression for Large Language Models","papers":24,"papersByYear":[24],"rel":null,"reviewed":true,"share":[0.6615],"size":27,"slope":null}],"totals":{"claims":[18312],"papers":[3628]},"years":[2026]}