大模型基准论文趋势 2023–2026

主要结论

• 大模型基准论文从 2023 年的 2 篇增长到 2026 年的 307 篇，三年间扩大到约 154 倍。
• NeurIPS 是发大模型基准论文最多的会议，2023–2026 年共 256 篇，占四个会议总量的 47%。
• 交叉方向里涨得最快的是AI 智能体：同时属于两个主题的论文从 2023 年的 1 篇增至 2026 年的 116 篇。
• 在「评测与基准」的 5 个主题里，它的 2026 年论文数排第 1；同组增长最快的模型即裁判是 27 倍，大模型基准是 154 倍。

代表论文

按年份和会议均匀挑选，方便快速看到每一届的样貌

• NeurIPS 2026A Free Lunch in LLM Compression: Revisiting Retraining after Pruning
• ICML 2026Controlled LLM Training on Spectral Sphere
• ICLR 2026Hey, That's My Model! Introducing Chain & Hash, An LLM Fingerprinting Technique
• CVPR 20263DrawAgent Teaching LLM to Draw in 3D with Early Contrastive
• NeurIPS 2025ACCO: Accumulate While You Communicate for Communication-Overlapped Sharded LLM Training
• ICML 2025Critical Tokens Matter: Token-Level Contrastive Estimation Enhances LLM’s Reasoning Capability
• ICLR 2025Catastrophic Failure of LLM Unlearning via Quantization
• CVPR 2025Period-LLM: Extending the Periodic Capability of Multimodal Large Language Model
• NeurIPS 2024Large Language Models' Expert-level Global History Knowledge Benchmark (HiST-LLM)
• ICML 2024Characterizing ResNet's Universal Approximation Capability
• ICLR 2024Faithful Explanations of Black-box NLP Models Using LLM-generated Counterfactuals
• CVPR 2024Koala: Key Frame-Conditioned Long Video-LLM
• NeurIPS 2023Quantizable Transformers: Removing Outliers by Helping Attention Heads Do Nothing
• NeurIPS 2026Accelerating LLM Pre-Training through Flat-Direction Dynamics Enhancement
• ICML 2026LLM Priors for ERM over Programs
• ICLR 2026How Learning Rate Decay Wastes Your Best Data in Curriculum-Based LLM Pretraining
• CVPR 20263D-VCD Hallucination Mitigation in 3D-LLM Embodied Agents through Visual Contrastive
• NeurIPS 2025BrainEC-LLM: Brain Effective Connectivity Estimation by Multiscale Mixing LLM
• ICML 2025Position: Medical Large Language Model Benchmarks Should Prioritize Construct Validity
• ICLR 2025Cheating Automatic LLM Benchmarks: Null Models Achieve High Win Rates

全部大模型基准论文

共 256 篇，第 1 / 6 页

1. NeurIPS 2026A Free Lunch in LLM Compression: Revisiting Retraining after Pruning
2. NeurIPS 2026A Retained-Signal Interface for LLM Watermark Robustness under Paraphrase
3. NeurIPS 2026Accelerating LLM Pre-Training through Flat-Direction Dynamics Enhancement
4. NeurIPS 2026Accelerating Long-Context LLM Prefill via Layer-wise Progressive Token Pruning in Local Deployment
5. NeurIPS 2026AccelEval: A Whole-Program Benchmark for LLM-Generated CPU-to-GPU Code Acceleration
6. NeurIPS 2026Adaptive Adversaries: A Multi-Turn, Multi-LLM Benchmark for LLM Agent Security
7. NeurIPS 2026ADKV: A Low-Overhead Adaptive Delta Quantization for KV Cache in LLM Inference
8. NeurIPS 2026Agent-ToM: Learning to Monitor Autonomous LLM Agents via Theory-of-Mind Reasoning
9. NeurIPS 2026AgentAbstain: Do LLM Agents Know When Not to Act?
10. NeurIPS 2026Aligning Language Model Benchmarks with Pairwise Preferences
11. NeurIPS 2026Alignment Dynamics in LLM Fine-Tuning
12. NeurIPS 2026Are LLM Safety Judges Policy-Invariant? A Three-Principle Stress-Test
13. NeurIPS 2026Are Multimodal Benchmarks Really Useful? Item-Level Multimodal Benchmark Diagnosis via Structure-Response Co-Calibration
14. NeurIPS 2026Ask in the Crowd: Differentially Private LLM Inference via Dummy-Augmented Shuffling
15. NeurIPS 2026ASPI: Seeking Ambiguity Clarification Amplifies Prompt Injection Vulnerability in LLM Agents
16. NeurIPS 2026AutoDataBench: How Far Are LLM Agents from Autonomously Engineering Post-Training Data Pipelines?
17. NeurIPS 2026BELLS-O: Evaluating the Operational Trade-offs of LLM Supervision Systems
18. NeurIPS 2026Benchmarks Are Not Atomic: Composition-Aware LLM Evaluation using BenchHub
19. NeurIPS 2026Beyond Linear Activation Steering: Invertible Latent Transformations for Controlling LLM Behavior
20. NeurIPS 2026BLADE: Scalable Bi-level Adaptive Data Selection for LLM Training
21. NeurIPS 2026Block-OBS-GS: Exact Per-Block Joint Brain Surgery with Gauss–Seidel Refinement for LLM Pruning
22. NeurIPS 2026BMAttn: Block-Aligned Mixed-Precision Attention Quantization for LLM Inference
23. NeurIPS 2026Boosting LLM Reasoning via Human-Inspired Reward Shaping
24. NeurIPS 2026Brittlebench: Quantifying LLM robustness via prompt sensitivity
25. NeurIPS 2026Calibration Is Not Control: Intervention Advantage for LLM-Agent Oversight
26. NeurIPS 2026Can an LLM Reason Like a Lawyer? Benchmarking the ability of LLMs to map the facts of a case to the elements of the applicable legal rule
27. NeurIPS 2026CapTrack: Multifaceted Evaluation of Forgetting in LLM Post-Training
28. NeurIPS 2026Capturing LLM Capabilities via Evidence-Calibrated Query Clustering
29. NeurIPS 2026Causal Benchmarks for Multimodal AI Should Measure Categorical Difference, Not Capability Gap
30. NeurIPS 2026CLIOPATRA: Extracting Private Information from LLM Insights
31. NeurIPS 2026COMPASS: Composable Policy-Amortized Structured Search for LLM-Based Optimization Modeling
32. NeurIPS 2026Confidence Estimation via Decoupled Smoothing for Dynamic LLM Routing and Aggregation
33. NeurIPS 2026Consolidating Rewarded Perturbations for LLM Post-Training
34. NeurIPS 2026ContractBench: Can LLM Agents Preserve Observation Contracts?
35. NeurIPS 2026Contrastive Representation Shaping for LLM Unlearning
36. NeurIPS 2026Correlation-Aware Contextual Bandits with Surrogate Rewards for LLM Routing
37. NeurIPS 2026CurveRL: Principled Distribution-Aware Context Reweighting for LLM Reasoning
38. NeurIPS 2026DAPS: Dependency-Aware Premise Selection for LLM Theorem Proving
39. NeurIPS 2026Decentralized Aggregation of LLM Predictions via Wagering Mechanisms
40. NeurIPS 2026Demystifying Numerical Errors in LLM Inference: Achieving Reproducible Inference for Mission-Critical Tasks with HEAL
41. NeurIPS 2026DICE: Decoupling Capability from Intervention Necessity in LLM Tutoring
42. NeurIPS 2026Differential Item Functioning as an Item-Level Diagnostic for LLM Benchmarks
43. NeurIPS 2026DistDebug-Bench: Can LLM Agents Diagnose Bugs in Distributed Systems?
44. NeurIPS 2026Don't Always Pick the Highest-Performing Model: An Information Theoretic View of LLM Ensemble Selection
45. NeurIPS 2026Efficient Evaluation of LLM Performance with Statistical Guarantees
46. NeurIPS 2026Efficient Gradient-Aware Asynchronous Reinforcement Learning for LLM Post-Training
47. NeurIPS 2026ENGINE: Endogenous Variational MultiScale Optimization for Zeroth-Order LLM Fine-Tuning
48. NeurIPS 2026Evaluating Physical Reasoning in LLM Agents Requires Construction Benchmarks
49. NeurIPS 2026Every Bit, Everywhere, All At Once: A Binomial Multibit LLM Watermark
50. NeurIPS 2026Fast-dLLM++: Fr\'{e}chet Profile Decoding for Faster Diffusion LLM Inference

常见问题

2023–2026 年四大会议一共有多少篇大模型基准论文？

共 544 篇，按年份分别是 2023 年 2 篇、2024 年 61 篇、2025 年 174 篇、2026 年 307 篇。

NeurIPS 2026 有多少篇大模型基准论文？

NeurIPS 2026 有 138 篇大模型基准论文，上一届是 92 篇。

ICLR 2026 有多少篇大模型基准论文？

ICLR 2026 有 103 篇大模型基准论文，上一届是 50 篇。

ICML 2026 有多少篇大模型基准论文？

ICML 2026 有 60 篇大模型基准论文，上一届是 24 篇。

CVPR 2026 有多少篇大模型基准论文？

CVPR 2026 有 6 篇大模型基准论文，上一届是 8 篇。

哪个会议的大模型基准论文最多？

NeurIPS，2023–2026 年共 256 篇。