NeurIPS 2026 accepted papers on Large language models
957 of the 6,080 papers accepted at NeurIPS 2026 (15.7%) are on Large language models, compared with 16.6% at NeurIPS 2025. They fall into 7 topics.
See NeurIPS 2026 in the interactive map
Working on something in this area? Describe your idea in scime atlas to see the closest accepted papers by meaning and how crowded the spot has become.
Topics and papers
Reasoning and chain of thought · 246 papers
- Beyond Semantics: The Unreasonable Effectiveness of Reasonless Intermediate Tokens (1 citation)
- Evaluating Test-Time Scaling of General LLM Agents (1 citation)
- Recursive Language Models (1 citation)
- A Communication-Theoretic Framework for LLM Agents: Cost-Aware Adaptive Reliability
- Agentic Multi-Turn Reasoning: A Fairness Approach
- An Enigma of Artificial Reason: Investigating the Production-Evaluation Gap in Large Reasoning Models
- Anti-Self-Distillation for Reasoning RL via Pointwise Mutual Information
- ARC-Encoder: learning compressed text representations for large language models
- Beyond What Seems Necessary: Hidden Gains from Scaling Training-Time Reasoning Length under Outcome Supervision
- BLADE: Scalable Bi-level Adaptive Data Selection for LLM Training
- Boosting Inference with Guided Reasoning: Stochastic Exploration for Recursive Models
- Boosting LLM Reasoning via Human-Inspired Reward Shaping
and 234 more in the interactive map.
Language, safety and interpretability · 226 papers
- A Bitter Lesson for Data Filtering
- A Single Neuron Is Sufficient to Bypass Safety Alignment in Large Language Models
- A Theory of Time-Sensitive Language Generation: Sparse Hallucination Beats Mode Collapse
- A Theory of Training Profit-Optimal LLMs
- Agnostic Language Identification and Generation
- Aligning Language Models with Selective Prediction
- Alignment Imprint: Zero-Shot AI-Generated Text Detection via Provable Preference Discrepancy
- Allocate Marginal Reviews to Borderline Papers Using LLM Comparative Ranking
- An Empirical Study on Noisy Data and LLM Pretraining Loss Divergence
- ArcMark: Distortion-Free Multi-Byte LLM Watermark via Optimal Transport
- Artificial Aphasias in Lesioned Language Models
- BBOWP-Bench: Evaluating LLMs on Black-Box Optimization Word Problems
and 214 more in the interactive map.
Agents, code and math · 157 papers
- MCP-Atlas: A Large-Scale Benchmark for Tool-Use Competency with Real MCP Servers (1 citation)
- Natural Language Actor-Critic: Scalable Off-Policy Learning in Language Space (1 citation)
- Reinforcement World Model Learning for LLM-based Agents (1 citation)
- \$OneMillion-Bench: How Far are Language Agents from Human Experts?
- ADAPT: Hybrid Prompt Optimization for LLM Feature Visualization
- Automated Reformulation of Robust Optimization via Memory-Augmented Large Language Models
- Can Agents Price a Reaction? Evaluating LLMs on Chemical Cost Reasoning
- Can Large Language Models Develop Gambling Addiction?
- CAREBench: Evaluating LLMs' Emotion Understanding by Assessing Cognitive Appraisal Reasoning
- CircuitSeer: Mining High-Quality Data by Probing Mathematical Reasoning Circuits in LLMs
- Coherence Mechanisms for Provable Self-Improvement
- CUDABench: Benchmarking LLMs for Text-to-CUDA Generation
and 145 more in the interactive map.
AI and society · 113 papers
- Working with AI: Measuring the Applicability of Generative AI to Occupations (4 citations)
- No One Knows the State-of-the-Art in Geospatial Foundation Models (2 citations)
- AI Evaluation Should Require Standardized Item-Level Data Releases (1 citation)
- Defining Operational Conditions for Safety-Critical AI-Based Systems from Data (1 citation)
- Explainability matters: The effect of liability rules on the healthcare sector (1 citation)
- Watermarking Without Standards Is Not AI Governance (1 citation)
- Adaptive auditing of AI systems with anytime-valid guarantees
- Agentic AI Scientists Are Not Built For Autonomous Scientific Discovery
- AI-Mediated Communication Can Steer Collective Opinion
- Auditing Sabotage Bench: A Benchmark for Detecting and Fixing Research Sabotage in ML Codebases
- BehaviorBench: Modeling Real-World User Decisions from Behavioral Traces
- Benchmark Health Index: A Systematic Framework for Benchmarking the Benchmarks of LLMs
and 101 more in the interactive map.
Retrieval and knowledge · 89 papers
- Medmarks: A Comprehensive Open-Source LLM Benchmark Suite for Medical Tasks (1 citation)
- A Semantic-Sampling Framework for Evaluating Calibration in Open-Ended Question Answering
- AgentHop: A Diagnostic Benchmark for Agentic Multi-Hop Scientific Question Answering
- AgenticOCR: Parsing Only What You Need for Efficient Retrieval-Augmented Generation
- Ask, Answer, and Detect: Role-Playing LLMs for Personality Detection with Question-Conditioned Mixture-of-Experts
- Autofocus Retrieval: An Effective Pipeline for Multi-Hop Question Answering With Semi-Structured Knowledge
- Beyond RAG for Agent Memory: Retrieval by Decoupling and Aggregation
- Beyond Raw Context Transfer: Representation-based Federated Retrieval-Augmented Generation
- Beyond Semantic Similarity: Rethinking Retrieval for Agentic Search via Direct Corpus Interaction
- Bian Que: An Agentic Framework with Flexible Skill Arrangement for Online System Operations
- Can Language Models Actually Retrieve In-Context? Drowning in Documents at Million Token Scale
- Can LLMs Take Retrieved Information with a Grain of Salt?
and 77 more in the interactive map.
Instruction and fine-tuning · 69 papers
- Alignment Dynamics in LLM Fine-Tuning
- AttnDiff: Attention-based Differential Fingerprinting for Large Language Models
- Back to Blackwell: Closing the Loop on Intransitivity in Multi-Objective Preference Fine-Tuning
- Beyond Parameter Arithmetic: Sparse Complementary Fusion for Distribution-Aware Model Merging
- Comparing Explanations is not Enough,Explain the Change: New Standards are Needed to Explain Behavioral Shifts in Large Language Models
- Crafting Reversible SFT Behaviors in Large Language Models
- Displacement Geometry Captures Platonic Shared Reality Across Models and Modalities
- Dominant-Layer ZO: A Single Layer Dominates Zeroth-Order Fine-Tuning of LLMs
- Efficiently Aligning Draft Models via Parameter- and Data-Efficient Adaptation
- ESSAM: A Novel Competitive Evolution Strategies Approach to Reinforcement Learning for Memory Efficient LLMs Fine-Tuning
- Evolution Fine-Tuning: Learning to Discover Across 371 Optimization Tasks
- Expected Harm: Rethinking Safety Evaluation of (Mis)Aligned LLMs
and 57 more in the interactive map.
Alignment and preferences · 57 papers
- A Unified Perturbation Framework for Analyzing Leaderboard Stability and Manipulation
- AdvJudge-Zero: Binary Decision Flips in LLM-as-a-Judge via Adversarial Control Tokens
- Aligning Language Model Benchmarks with Pairwise Preferences
- Bayesian Preference Learning for Test-Time Steerable Reward Models
- BSO: Safety Alignment Is Density Ratio Matching
- Curriculum Learning for Safety Alignment
- Distributionally Robust Listwise Preference Optimization
- Distributionally Robust Token Optimization in RLHF
- From “Weak” Signals to Strong Models: Preference Delta Aggregation with LoRA Merging
- How Does Personalized Memory Shape LLM Behavior? Benchmarking Rational Preference Utilization in Personalized Assistants
- Inference Time Nash Alignment
- INSPO : Unlocking Intrinsic Self-Reflection for LLM Preference Optimization
and 45 more in the interactive map.