LLM Benchmarks papers, 2023–2026

Key findings

• LLM Benchmarks went from 2 papers in 2023 to 307 in 2026, roughly 154x in three years.
• NeurIPS publishes the most LLM Benchmarks work: 256 papers from 2023 to 2026, 47% of the four-venue total.
• The fastest-growing overlap is with AI Agents: papers tagged with both rose from 1 in 2023 to 116 in 2026.
• Among the 5 topics under Evaluation and Benchmarks, it ranks #1 by 2026 paper count; the fastest grower in the group, LLM-as-Judge, is up 27x versus 154x here.

Representative papers

Spread across years and venues so every edition shows up

• 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

All LLM Benchmarks papers

307 papers, page 4 of 7

151. ICML 2026Scheduling LLM Inference with Uncertainty-Aware Output Length Predictions
152. ICML 2026SetPO: Set-Level Policy Optimization for Diversity-Preserving LLM Reasoning
153. ICML 2026Single-Rollout Hidden-State Dynamics for Training-Free RLVR Data Selection
154. ICML 2026Skill-Pro: Learning Reusable Skills from Experience via Non-Parametric PPO for LLM Agents
155. ICML 2026SPR-RAFT: Parameter-Efficient Regression-Aware Fine-Tuning for Biomedical LLM Regression
156. ICML 2026Spurious Rewards: Rethinking Training Signals in RLVR
157. ICML 2026The Invisible Lottery: How Subtle Cues Steer Algorithm Choice in LLM Code Generation
158. ICML 2026Think Twice Before You Act: Protecting LLM Agents Against Tool Description Poisoning via Isolated Planning
159. ICML 2026Tiny Brains, Giant Impact: Uncovering the Keystone Neurons of LLM with Just a Few Prompts
160. ICML 2026Token-Level LLM Collaboration via FusionRoute
161. ICML 2026Toward Cybersecurity-Expert Small Language Models
162. ICML 2026Training-Free Hashing-Based Attention via Binary Principal Components
163. ICML 2026TwinQuant: Learnable Subspace Decomposition for 4-Bit LLM Quantization
164. ICML 2026UltraHorizon: Benchmarking LLM-Agent Capabilities in Ultra Long-Horizon Scenarios
165. ICML 2026Unveiling the Potential of Quantization with MXFP4: Strategies for Quantization Error Reduction
166. ICML 2026Utility-Diversity Aware Online Batch Selection for LLM Supervised Fine-tuning
167. ICML 2026What Do Agents Learn from Trajectory-SFT: Semantics or Interfaces?
168. ICML 2026When Single Answer Is Not Enough: Rethinking Single-Step Retrosynthesis Benchmarks for LLMs
169. ICML 2026Zero Sum SVD: Balancing Loss Sensitivity for Low Rank LLM Compression
170. NeurIPS 2026A Free Lunch in LLM Compression: Revisiting Retraining after Pruning
171. NeurIPS 2026A Retained-Signal Interface for LLM Watermark Robustness under Paraphrase
172. NeurIPS 2026Accelerating LLM Pre-Training through Flat-Direction Dynamics Enhancement
173. NeurIPS 2026Accelerating Long-Context LLM Prefill via Layer-wise Progressive Token Pruning in Local Deployment
174. NeurIPS 2026AccelEval: A Whole-Program Benchmark for LLM-Generated CPU-to-GPU Code Acceleration
175. NeurIPS 2026Adaptive Adversaries: A Multi-Turn, Multi-LLM Benchmark for LLM Agent Security
176. NeurIPS 2026ADKV: A Low-Overhead Adaptive Delta Quantization for KV Cache in LLM Inference
177. NeurIPS 2026Agent-ToM: Learning to Monitor Autonomous LLM Agents via Theory-of-Mind Reasoning
178. NeurIPS 2026AgentAbstain: Do LLM Agents Know When Not to Act?
179. NeurIPS 2026Aligning Language Model Benchmarks with Pairwise Preferences
180. NeurIPS 2026Alignment Dynamics in LLM Fine-Tuning
181. NeurIPS 2026Are LLM Safety Judges Policy-Invariant? A Three-Principle Stress-Test
182. NeurIPS 2026Are Multimodal Benchmarks Really Useful? Item-Level Multimodal Benchmark Diagnosis via Structure-Response Co-Calibration
183. NeurIPS 2026Ask in the Crowd: Differentially Private LLM Inference via Dummy-Augmented Shuffling
184. NeurIPS 2026ASPI: Seeking Ambiguity Clarification Amplifies Prompt Injection Vulnerability in LLM Agents
185. NeurIPS 2026AutoDataBench: How Far Are LLM Agents from Autonomously Engineering Post-Training Data Pipelines?
186. NeurIPS 2026BELLS-O: Evaluating the Operational Trade-offs of LLM Supervision Systems
187. NeurIPS 2026Benchmarks Are Not Atomic: Composition-Aware LLM Evaluation using BenchHub
188. NeurIPS 2026Beyond Linear Activation Steering: Invertible Latent Transformations for Controlling LLM Behavior
189. NeurIPS 2026BLADE: Scalable Bi-level Adaptive Data Selection for LLM Training
190. NeurIPS 2026Block-OBS-GS: Exact Per-Block Joint Brain Surgery with Gauss–Seidel Refinement for LLM Pruning
191. NeurIPS 2026BMAttn: Block-Aligned Mixed-Precision Attention Quantization for LLM Inference
192. NeurIPS 2026Boosting LLM Reasoning via Human-Inspired Reward Shaping
193. NeurIPS 2026Brittlebench: Quantifying LLM robustness via prompt sensitivity
194. NeurIPS 2026Calibration Is Not Control: Intervention Advantage for LLM-Agent Oversight
195. 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
196. NeurIPS 2026CapTrack: Multifaceted Evaluation of Forgetting in LLM Post-Training
197. NeurIPS 2026Capturing LLM Capabilities via Evidence-Calibrated Query Clustering
198. NeurIPS 2026Causal Benchmarks for Multimodal AI Should Measure Categorical Difference, Not Capability Gap
199. NeurIPS 2026CLIOPATRA: Extracting Private Information from LLM Insights
200. NeurIPS 2026COMPASS: Composable Policy-Amortized Structured Search for LLM-Based Optimization Modeling

FAQ

How many LLM Benchmarks papers were published at NeurIPS, ICLR, ICML and CVPR from 2023 to 2026?

544 in total: 2 in 2023, 61 in 2024, 174 in 2025, 307 in 2026.

How many LLM Benchmarks papers are at NeurIPS 2026?

NeurIPS 2026 has 138 LLM Benchmarks papers, versus 92 the year before.

How many LLM Benchmarks papers are at ICLR 2026?

ICLR 2026 has 103 LLM Benchmarks papers, versus 50 the year before.

How many LLM Benchmarks papers are at ICML 2026?

ICML 2026 has 60 LLM Benchmarks papers, versus 24 the year before.

How many LLM Benchmarks papers are at CVPR 2026?

CVPR 2026 has 6 LLM Benchmarks papers, versus 8 the year before.

Which conference publishes the most LLM Benchmarks papers?

NeurIPS, with 256 papers from 2023 to 2026.