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 174 篇，第 2 / 4 页

51. ICLR 2025SpinQuant: LLM Quantization with Learned Rotations
52. ICLR 2025SqueezeAttention: 2D Management of KV-Cache in LLM Inference via Layer-wise Optimal Budget
53. ICLR 2025StringLLM: Understanding the String Processing Capability of Large Language Models
54. ICLR 2025Style Outweighs Substance: Failure Modes of LLM Judges in Alignment Benchmarking
55. ICLR 2025TODO: Enhancing LLM Alignment with Ternary Preferences
56. ICLR 2025ToolACE: Winning the Points of LLM Function Calling
57. ICLR 2025Towards a Theoretical Understanding of Synthetic Data in LLM Post-Training: A Reverse-Bottleneck Perspective
58. ICLR 2025Trajectory-LLM: A Language-based Data Generator for Trajectory Prediction in Autonomous Driving
59. ICML 2025CateKV: On Sequential Consistency for Long-Context LLM Inference Acceleration
60. ICML 2025Critical Tokens Matter: Token-Level Contrastive Estimation Enhances LLM’s Reasoning Capability
61. ICML 2025Demystifying Singular Defects in Large Language Models
62. ICML 2025Exploring Criteria of Loss Reweighting to Enhance LLM Unlearning
63. ICML 2025FlatQuant: Flatness Matters for LLM Quantization
64. ICML 2025From Crowdsourced Data to High-quality Benchmarks: Arena-Hard and Benchbuilder Pipeline
65. ICML 2025From Jack of All Trades to Master of One: Specializing LLM-based Autoraters to a Test Set
66. ICML 2025La RoSA: Enhancing LLM Efficiency via Layerwise Rotated Sparse Activation
67. ICML 2025Large Language-Geometry Model: When LLM meets Equivariance
68. ICML 2025LongRoPE2: Near-Lossless LLM Context Window Scaling
69. ICML 2025Memorization Sinks: Isolating Memorization during LLM Training
70. ICML 2025Minerva: A Programmable Memory Test Benchmark for Language Models
71. ICML 2025Overestimation in LLM Evaluation: A Controlled Large-Scale Study on Data Contamination’s Impact on Machine Translation
72. ICML 2025Position: Don't Use the CLT in LLM Evals With Fewer Than a Few Hundred Datapoints
73. ICML 2025Position: Medical Large Language Model Benchmarks Should Prioritize Construct Validity
74. ICML 2025Position: Principles of Animal Cognition to Improve LLM Evaluations
75. ICML 2025Position: Scaling LLM Agents Requires Asymptotic Analysis with LLM Primitives
76. ICML 2025Position: The Most Expensive Part of an LLM *should* be its Training Data
77. ICML 2025RBench: Graduate-level Multi-disciplinary Benchmarks for LLM & MLLM Complex Reasoning Evaluation
78. ICML 2025RE-Bench: Evaluating Frontier AI R&D Capabilities of Language Model Agents against Human Experts
79. ICML 2025Reward-Augmented Data Enhances Direct Preference Alignment of LLMs
80. ICML 2025RZ-NAS: Enhancing LLM-guided Neural Architecture Search via Reflective Zero-Cost Strategy
81. ICML 2025Sketch to Adapt: Fine-Tunable Sketches for Efficient LLM Adaptation
82. ICML 2025TruthFlow: Truthful LLM Generation via Representation Flow Correction
83. NeurIPS 2025$Q\sharp$: Provably Optimal Distributional RL for LLM Post-Training
84. NeurIPS 2025Accelerating Block Coordinate Descent for LLM Finetuning via Landscape Expansion
85. NeurIPS 2025ACCO: Accumulate While You Communicate for Communication-Overlapped Sharded LLM Training
86. NeurIPS 2025Accurate KV Cache Eviction via Anchor Direction Projection for Efficient LLM Inference
87. NeurIPS 2025Act Only When It Pays: Efficient Reinforcement Learning for LLM Reasoning via Selective Rollouts
88. NeurIPS 2025AgentRecBench: Benchmarking LLM Agent-based Personalized Recommender Systems
89. NeurIPS 2025Antidistillation Sampling
90. NeurIPS 2025AutoJudge: Judge Decoding Without Manual Annotation
91. NeurIPS 2025BenchmarkCards: Standardized Documentation for Large Language Model Benchmarks
92. NeurIPS 2025Beyond Chemical QA: Evaluating LLM's Chemical Reasoning with Modular Chemical Operations
93. NeurIPS 2025BLEUBERI: BLEU is a surprisingly effective reward for instruction following
94. NeurIPS 2025Bohdi: Heterogeneous LLM Fusion with Automatic Data Exploration
95. NeurIPS 2025BrainEC-LLM: Brain Effective Connectivity Estimation by Multiscale Mixing LLM
96. NeurIPS 2025C-SEO Bench: Does Conversational SEO Work?
97. NeurIPS 2025Can Dependencies Induced by LLM-Agent Workflows Be Trusted?
98. NeurIPS 2025Causal LLM Routing: End-to-End Regret Minimization from Observational Data
99. NeurIPS 2025ComPO: Preference Alignment via Comparison Oracles
100. NeurIPS 2025ConnectomeBench: Can LLMs proofread the connectome?

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。