ICML 2026 · 全部收录(566 篇)¶
仅列出通过预过滤且已打分的论文;分数是「评测方法学价值」,不是会议录用质量。 返回总览
| # | 论文 | 轨道 | 分数 | 链接 |
|---|---|---|---|---|
| 151 | ProEval: Proactive Failure Discovery and Efficient Performance Estimation for Generative AI Evaluation | Main | 8 | 页 · PDF |
| 152 | ProtDBench: A Unified Benchmark of Protein Binder Design and Evaluation | Main | 8 | 页 · PDF |
| 153 | Query-efficient model evaluation using cached responses | Main | 8 | 页 · PDF |
| 154 | REAL: Regression-Aware Reinforcement Learning for LLM-as-a-Judge | Main | 8 | 页 · PDF |
| 155 | Recovering Policy-Induced Errors: Benchmarking and Trajectory Synthesis for Robust GUI Agents | Main | 8 | 页 · PDF |
| 156 | Rethinking GNNs and Missing Features: Challenges, Evaluation and a Robust Solution | Main | 8 | 页 · PDF |
| 157 | Rethinking Multimodal Time-Series Forecasting Evaluation | Main | 8 | 页 · PDF |
| 158 | ReVSI: Rebuilding Visual Spatial Intelligence Evaluation for Accurate Assessment of VLM 3D Reasoning | Main | 8 | 页 · PDF |
| 159 | Reward Hacking Benchmark: Measuring Exploits in LLM Agents with Tool Use | Main | 8 | 页 · PDF |
| 160 | Routing and Reasoned Evaluation with Large Language Models | Main | 8 | 页 · PDF |
| 161 | SAW-Bench: Learning Situated Awareness in the Real World | Main | 8 | 页 · PDF |
| 162 | Scaling Robot Policy Evaluation via Discrete Diffusion World Model | Main | 8 | 页 · PDF |
| 163 | SciAgentGym: Benchmarking Multi-Step Scientific Tool-Use in LLM Agents | Main | 8 | 页 · PDF |
| 164 | SCOPE and SCION: A Benchmark and an Auditable Reference Pipeline for Schema Induction and Fusion from Text | Main | 8 | 页 · PDF |
| 165 | SimulCost: A Cost-Aware Benchmark and Toolkit for Automating Physics Simulations with LLMs | Main | 8 | 页 · PDF |
| 166 | SPEED-Bench: A Unified and Diverse Benchmark for Speculative Decoding | Main | 8 | 页 · PDF |
| 167 | SWE-IF: Aligning Code Evaluation with Human Preference | Main | 8 | 页 · PDF |
| 168 | SWE-Milestone: Evaluating AI Agents on Continuous Software Evolution | Main | 8 | 页 · PDF |
| 169 | TadA-Bench: A Million-Variant Benchmark for Future-Round Discovery Toward Agentic Protein Engineering | Main | 8 | 页 · PDF |
| 170 | TestExplora: Benchmarking LLMs for Proactive Bug Discovery via Repository-Level Test Generation | Main | 8 | 页 · PDF |
| 171 | The ACUTE Protocol: Operationalizing Language Model Activations for Better Calibration, Utility, and Trust | Main | 8 | 页 · PDF |
| 172 | The Decrypto Benchmark for Multi-Agent Reasoning and Theory of Mind | Main | 8 | 页 · PDF |
| 173 | Thinking in Structures: Evaluating Spatial Intelligence in Constraint-Governed Spaces | Main | 8 | 页 · PDF |
| 174 | TokSuite: Measuring the Impact of Tokenizer Choice on Language Model Behavior | Main | 8 | 页 · PDF |
| 175 | Towards Trustworthy Video Anomaly Understanding: A Class-Guided Chain-of-Evaluation Metric and An Anomaly-focused Meta-Benchmark | Main | 8 | 页 · PDF |
| 176 | UltraHorizon: Benchmarking Agent Capabilities in Ultra Long-Horizon Scenarios | Main | 8 | 页 · PDF |
| 177 | Uncovering Competency Gaps in Large Language Models and Their Benchmarks | Main | 8 | 页 · PDF |
| 178 | UniCode: Augmenting Evaluation for Code Reasoning | Main | 8 | 页 · PDF |
| 179 | Unlearning Isn’t Forgetting: Revealing Hidden Leakage in Class Unlearning Evaluations | Main | 8 | 页 · PDF |
| 180 | Unsafer in Many Turns: Benchmarking and Defending Multi-Turn Safety Risks in Tool-Using Agents | Main | 8 | 页 · PDF |
| 181 | VenusBench-Mobile: A Challenging and User-Centric Benchmark for Mobile GUI Agents with Capability Diagnostics | Main | 8 | 页 · PDF |
| 182 | VLA-Arena: An Open-Source Framework for Benchmarking Vision-Language-Action Models | Main | 8 | 页 · PDF |
| 183 | VocSim A Training-free Benchmark for Zero-shot Content Identity in Single-source Audio | Main | 8 | 页 · PDF |
| 184 | When Single Answer Is Not Enough: Rethinking Single-Step Retrosynthesis Benchmarks for LLMs | Main | 8 | 页 · PDF |
| 185 | Who Evaluates AI’s Social Impacts? Mapping Coverage and Gaps in First and Third Party Evaluations | Main | 8 | 页 · PDF |
| 186 | WorldTravel: A Realistic Multimodal Travel-Planning Benchmark with Tightly Coupled Constraints | Main | 8 | 页 · PDF |
| 187 | \(τ\)-Voice: Benchmarking Full-Duplex Voice Agents on Real-World Domains | Main | 7 | 页 · PDF |
| 188 | A Behavioural and Representational Evaluation of Goal-Directedness in Language Model Agents | Main | 7 | 页 · PDF |
| 189 | A Computational Framework for Evaluating Human-likeness in LLMs’ Open-ended Human Behaviors | Main | 7 | 页 · PDF |
| 190 | Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers | Main | 7 | 页 · PDF |
| 191 | AgentHijack: Benchmarking Computer Use Agent Robustness to Common Environment Corruptions | Main | 7 | 页 · PDF |
| 192 | AgentLAB: Benchmarking LLM Agents against Long-Horizon Attacks | Main | 7 | 页 · PDF |
| 193 | AgentWebBench: Benchmarking Multi-Agent Coordination in Agentic Web | Main | 7 | 页 · PDF |
| 194 | AMA-Bench: Evaluating Long-Horizon Memory for Agentic Applications | Main | 7 | 页 · PDF |
| 195 | APE-Bench: Evaluating Automated Proof Engineering for Formal Math Libraries | Main | 7 | 页 · PDF |
| 196 | ArenaRL: Scaling RL for Open-Ended Agents via Tournament-based Relative Ranking | Main | 7 | 页 · PDF |
| 197 | Assistive Prompt Mediation: Evaluating Language Models Under Accessibility Constraints | Main | 7 | 页 · PDF |
| 198 | AtelierEval: Agentic Evaluation of Humans & LLMs as Text-to-Image Prompters | Main | 7 | 页 · PDF |
| 199 | AutoControl Arena: Synthesizing Executable Test Environments for Frontier AI Risk Evaluation | Main | 7 | 页 · PDF |
| 200 | AVI-Bench: Toward Human-like Audio-Visual Intelligence of Omni-MLLMs | Main | 7 | 页 · PDF |
| 201 | BEDTime: A Unified Benchmark for Automatically Describing Time Series | Main | 7 | 页 · PDF |
| 202 | Benchmarking the Limits of In-Context Reinforcement Learning for Ad-Hoc Teamwork | Main | 7 | 页 · PDF |
| 203 | Benchmarking the Scientific Mind: A Pathology-Derived Biomedical VQA Benchmark for Complex Scientific Reasoning | Main | 7 | 页 · PDF |
| 204 | BESPOKE: Benchmark for Search-Augmented Large Language Model Personalization via Diagnostic Feedback | Main | 7 | 页 · PDF |
| 205 | BEST: Benchmarking Efficiency in Space and Time for LLM-Generated Code | Main | 7 | 页 · PDF |
| 206 | Beyond Problem Solving: UOJ-Bench for Evaluating Code Generation, Hacking, and Repair in Competitive Programming | Main | 7 | 页 · PDF |
| 207 | Beyond Scalars: Evaluating and Understanding LLM Reasoning via Geometric Progress and Stability | Main | 7 | 页 · PDF |
| 208 | Beyond Theorem Proving: Formulation, Framework and Benchmark for Formal Problem-Solving | Main | 7 | 页 · PDF |
| 209 | BioAgent Bench: An AI Agent Evaluation Suite for Bioinformatics | Main | 7 | 页 · PDF |
| 210 | BioProBench: A Corpus and Benchmark for Biological Protocol Reasoning in Autonomous Science | Main | 7 | 页 · PDF |
| 211 | BizFinBench.v2: Towards Reliable LLMs in Finance via Real-User Data and Offline/Online Bilingual Evaluation | Main | 7 | 页 · PDF |
| 212 | BuildArena: A Physics-Aligned Interactive Benchmark of LLMs for Engineering Construction | Main | 7 | 页 · PDF |
| 213 | Can LLM Agents Stick to the Script? A Benchmark for Long-Horizon Consistency in Interactive Narratives | Main | 7 | 页 · PDF |
| 214 | Can LLMs Reason Structurally? Benchmarking via the lens of Data Structures | Main | 7 | 页 · PDF |
| 215 | CaP-X: A Framework for Benchmarking and Improving Coding Agents for Robot Manipulation | Main | 7 | 页 · PDF |
| 216 | CATArena: Evaluating Evolutionary Capabilities of Code Agents via Iterative Tournaments | Main | 7 | 页 · PDF |
| 217 | CausalGame: Benchmarking Causal Thinking of LLM Agents in Games | Main | 7 | 页 · PDF |
| 218 | Characterizing, Evaluating, and Optimizing Complex Reasoning | Main | 7 | 页 · PDF |
| 219 | CIRBench: Evaluating Large Language Models as LLVM IR Optimizers | Main | 7 | 页 · PDF |
| 220 | CMI-RewardBench: Evaluating Music Reward Models with Compositional Multimodal Instruction | Main | 7 | 页 · PDF |
| 221 | CoDA-Bench: Can Code Agents Handle Data-Intensive Tasks? | Main | 7 | 页 · PDF |
| 222 | CodeClash: Benchmarking Goal-Oriented Software Engineering | Main | 7 | 页 · PDF |
| 223 | CollabBench: Benchmarking and Unleashing Collaborative Ability of LLMs with Diverse Players via Proactive Engagement | Main | 7 | 页 · PDF |
| 224 | Conditional Coverage Diagnostics for Conformal Prediction | Main | 7 | 页 · PDF |
| 225 | Conformal Reliability: A New Evaluation Metric for Conditional Generation | Main | 7 | 页 · PDF |
| 226 | ConTSG-Bench: A Unified Benchmark for Conditional Time Series Generation | Main | 7 | 页 · PDF |
| 227 | CoopEval: Benchmarking Cooperation-Sustaining Mechanisms and LLM Agents in Social Dilemmas | Main | 7 | 页 · PDF |
| 228 | Data Difficulty and the Generalization–Extrapolation Tradeoff in LLM Fine-Tuning | Main | 7 | 页 · PDF |
| 229 | Deep Learning for BioImaging: What Are We Really Learning? | Main | 7 | 页 · PDF |
| 230 | Demystifying LLM-as-a-Judge: Analytically Tractable Model for Inference-Time Scaling | Main | 7 | 页 · PDF |
| 231 | Distinguishing Imitation Error from Intrinsic Motion Learning Difficulty | Main | 7 | 页 · PDF |
| 232 | DLEBench: Evaluating Small-scale Object Editing Ability for Instruction-based Image Editing Model | Main | 7 | 页 · PDF |
| 233 | DocHop: Benchmarking Out-of-domain Multi-hop Reasoning in Information-Dense Documents | Main | 7 | 页 · PDF |
| 234 | DV-World: Benchmarking Data Visualization Agents in Real-World Scenarios | Main | 7 | 页 · PDF |
| 235 | EEG-FM-Bench: A Comprehensive Benchmark for the Systematic Evaluation and Diagnostic Analyses of EEG Foundation Models | Main | 7 | 页 · PDF |
| 236 | ERGeoBench: A Comprehensive Benchmark for Embodied Reasoning and Geo-localization in Multimodal Large Language Models | Main | 7 | 页 · PDF |
| 237 | Evaluating and Explaining Prompt Sensitivity of LLMs Using Interactions | Main | 7 | 页 · PDF |
| 238 | Evaluating Contextual Illegality: AI Compliance in Corporate Law Scenarios | Main | 7 | 页 · PDF |
| 239 | Evaluating Newtonian Mechanics in Video Generative Models with Real Physical Systems | Main | 7 | 页 · PDF |
| 240 | Evaluating Object-Centric Models beyond Object Discovery | Main | 7 | 页 · PDF |
| 241 | Evaluating Relational Reasoning in LLMs with REL | Main | 7 | 页 · PDF |
| 242 | EVMbench: Evaluating AI Agents on Smart Contract Security | Main | 7 | 页 · PDF |
| 243 | ExCyTIn-Bench: Evaluating LLM agents on Cyber Threat Investigation | Main | 7 | 页 · PDF |
| 244 | FakeWorld 1.0: An Omni-modal Benchmark for Fake Media and Content | Main | 7 | 页 · PDF |
| 245 | FiRE: Fine-grained Ranking Evaluation for Machine Translation | Main | 7 | 页 · PDF |
| 246 | Fix Before Search: Benchmarking Agentic Visual Query Pre-processing in Multimodal Retrieval-augmented Generation | Main | 7 | 页 · PDF |
| 247 | FLIP2: Expanding Protein Fitness Landscape Benchmarks for Real-World Machine Learning Applications | Main | 7 | 页 · PDF |
| 248 | FloorplanQA: A Benchmark for Spatial Reasoning in LLMs using Structured Representations | Main | 7 | 页 · PDF |
| 249 | From Evaluation to Design: Using Potential Energy Surface Smoothness Metrics to Guide ML Interatomic Potential Architectures | Main | 7 | 页 · PDF |
| 250 | FutureOmni: Evaluating Future Forecasting from Omni-Modal Context for Multimodal LLMs | Main | 7 | 页 · PDF |
| 251 | Graph Alignment for Benchmarking Graph Neural Networks and Learning Positional Encodings | Main | 7 | 页 · PDF |
| 252 | HEARTS: Benchmarking LLM Reasoning on Health Time Series | Main | 7 | 页 · PDF |
| 253 | How Far Can LLM Agents Reason with Tables? Benchmarking Multi-Turn Agentic Table Question Answering in the Wild | Main | 7 | 页 · PDF |
| 254 | How Powerful are LLMs in Generating Formal Program Specifications? | Main | 7 | 页 · PDF |
| 255 | How2Everything: Mining the Web for How-to Procedures to Evaluate and Improve LLMs | Main | 7 | 页 · PDF |
| 256 | ImpText: A Benchmark and Tool-Augmented Framework for Implicit Text Reasoning | Main | 7 | 页 · PDF |
| 257 | InfraRL: A Benchmark for Constrained Resource Allocation in Large-Scale Infrastructure Asset Management | Main | 7 | 页 · PDF |
| 258 | InteractBench: Benchmarking LLMs on Competitive Programming under Unrevealed Information | Main | 7 | 页 · PDF |
| 259 | InteractScience: Programmatic and Visually-Grounded Evaluation of Interactive Scientific Demonstration Code Generation | Main | 7 | 页 · PDF |
| 260 | Is Vibe Coding Safe? Benchmarking Vulnerability of Agent-Generated Code in Real-World Tasks | Main | 7 | 页 · PDF |
| 261 | It’s TIME: Towards the Next Generation of Time Series Forecasting Benchmarks | Main | 7 | 页 · PDF |
| 262 | KAGE-Bench: Fast Known-Axis Visual Generalization Evaluation for Reinforcement Learning | Main | 7 | 页 · PDF |
| 263 | Less Precise Can Be More Reliable: A Systematic Evaluation of Quantization’s Impact on VLMs Beyond Accuracy | Main | 7 | 页 · PDF |
| 264 | LoCoT2V-Bench: Benchmarking Long-Form and Complex Text-to-Video Generation | Main | 7 | 页 · PDF |
| 265 | MAS-Orchestra: Understanding and Improving Multi-Agent Reasoning Through Holistic Orchestration and Controlled Benchmarks | Main | 7 | 页 · PDF |
| 266 | MCP-Persona: Benchmarking LLM Agents on Real-World Personal Applications via Environment Simulation | Main | 7 | 页 · PDF |
| 267 | MEAL: A Benchmark for Continual Multi-Agent Reinforcement Learning | Main | 7 | 页 · PDF |
| 268 | Measuring and Mitigating Post-Hoc Rationalization in Reverse Chain-of-Thought Generation | Main | 7 | 页 · PDF |
| 269 | Measuring Intent Comprehension in LLMs | Main | 7 | 页 · PDF |
| 270 | MECAT: A Multi-Experts Constructed Benchmark for Fine-Grained Audio Understanding Tasks | Main | 7 | 页 · PDF |
| 271 | MemoryBench: A Benchmark for Memory and Continual Learning in LLM Systems | Main | 7 | 页 · PDF |
| 272 | MET-Bench: Multimodal Entity Tracking for Evaluating the Limitations of Vision-Language and Reasoning Models | Main | 7 | 页 · PDF |
| 273 | MiniAppBench: Evaluating the Shift from Text to Interactive HTML Responses in LLM-Powered Assistants | Main | 7 | 页 · PDF |
| 274 | MLUBench: A Benchmark for Lifelong Unlearning Evaluation in MLLMs | Main | 7 | 页 · PDF |
| 275 | MMClima: A Framework for Multimodal Climate Science Data and Evaluation | Main | 7 | 页 · PDF |
| 276 | MMKU-Bench: A Multimodal Update Benchmark for Diverse Visual Knowledge | Main | 7 | 页 · PDF |
| 277 | MORALISE: A Structured Benchmark for Moral Alignment in Visual Language Models | Main | 7 | 页 · PDF |
| 278 | MORE: A Multilingual Document Parsing Benchmark and Evaluation | Main | 7 | 页 · PDF |
| 279 | MultiHal: Multilingual Dataset for Knowledge-Graph Grounded Evaluation of LLM Hallucinations | Main | 7 | 页 · PDF |
| 280 | MVI-Bench: A Comprehensive Benchmark for Evaluating Robustness to Misleading Visual Inputs in LVLMs | Main | 7 | 页 · PDF |
| 281 | NL2Repo-Bench: Towards Long-Horizon Repository Generation Evaluation of Coding Agents | Main | 7 | 页 · PDF |
| 282 | One Probe Won’t Catch Them All: Towards Targeted Deception Detection | Main | 7 | 页 · PDF |
| 283 | Op-CAD: Benchmarking and Investigating Operation-oriented CAD Generation | Main | 7 | 页 · PDF |
| 284 | Out-of-Distribution Evaluation of Rule-Based and Strategic Reasoning in Chess Transformers | Main | 7 | 页 · PDF |
| 285 | PhoStream: Benchmarking Real-World Streaming for Omnimodal Assistants in Mobile Scenarios | Main | 7 | 页 · PDF |
| 286 | Plug-and-Play Benchmarking of Reinforcement Learning Algorithms for Large-Scale Flow Control | Main | 7 | 页 · PDF |
| 287 | Position: AI Evaluation Should Work With Humans | Main | 7 | 页 · PDF |
| 288 | Position: Code Benchmarks Should Prioritize Rigor, Reliability, and Reproducibility | Main | 7 | 页 · PDF |
| 289 | Position: Evaluation of ECG Representations Must Be Fixed | Main | 7 | 页 · PDF |
| 290 | Position: Mathematical DSL Benchmarks Need Executable Interface Contracts | Main | 7 | 页 · PDF |
| 291 | Position: Mechanisms for Aggregated Individual Reporting Should be Established for Post-Deployment Evaluation | Main | 7 | 页 · PDF |
| 292 | Position: Stop Chasing the C-index when Evaluating Survival Analysis Models | Main | 7 | 页 · PDF |
| 293 | Position: Stop Evaluating AI with Human Tests, Develop Principled, AI-specific Tests instead | Main | 7 | 页 · PDF |
| 294 | Position: Stop Using Culturally Biased Human Cognitive Benchmarks to Evaluate LLMs | Main | 7 | 页 · PDF |
| 295 | Position: Time-Series Foundation Models Require Explicit Domain-Level Benchmarks | Main | 7 | 页 · PDF |
| 296 | ProAct: A Benchmark and Multimodal Framework for Structure-Aware Proactive Response | Main | 7 | 页 · PDF |
| 297 | Proactive Defense Benchmark against Deepfake Generation | Main | 7 | 页 · PDF |
| 298 | ProMiSE: Protein Multi-State Evaluation Benchmark in Biological Contexts | Main | 7 | 页 · PDF |
| 299 | ProOPF: Benchmarking and Improving LLMs for Professional-Grade Power Systems Optimization Modeling | Main | 7 | 页 · PDF |
| 300 | QiMeng-LibBench: Benchmarking LLM Agents for Library-Scale Cross-Architecture Migration | Main | 7 | 页 · PDF |