AIME-25

AIME-25 is a benchmark dataset used to evaluate the performance of models in long-context reasoning tasks.

Papers using AIME-25 (30)

Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoning2025 · 1 citesTransformation-Augmented GRPO for Enhancing Exploration in Reasoning of Large Language Models2026GlimpRouter: Efficient Collaborative Inference by Glimpsing One Token of Thoughts2026VTC-R1: Vision-Text Compression for Efficient Long-Context Reasoning2026TriAttention: Efficient Long Reasoning with Trigonometric KV Compression2026From Reasoning Chains to Verifiable Subproblems: Curriculum Reinforcement Learning Enables Credit Assignment for LLM Reasoning2026Learn Hard Problems During RL with Reference Guided Fine-tuning2026SePO: Self-Evolving Prompt Agent for System Prompt Optimization2026Benchmarking EngGPT2-16B-A3B against Comparable Italian and International Open-source LLMs2026When to Think, When to Speak: Learning Disclosure Policies for LLM Reasoning2026Test-time Recursive Thinking: Self-Improvement without External Feedback2026Prompting Test-Time Scaling Is A Strong LLM Reasoning Data Augmentation2025RefCritic: Training Long Chain-of-Thought Critic Models with Refinement Feedback2025MoL-RL: Distilling Multi-Step Environmental Feedback into LLMs for Feedback-Independent Reasoning2025Skywork Open Reasoner 1 Technical Report2025RefCritic: Training Long Chain-of-Thought Critic Models with Refinement Feedback2025SAND-Math: Using LLMs to Generate Novel, Difficult and Useful Mathematics Questions and Answers2025Beyond Pass@1: Self-Play with Variational Problem Synthesis Sustains RLVR2025DCPO: Dynamic Clipping Policy Optimization2025Light-R1: Curriculum SFT, DPO and RL for Long COT from Scratch and Beyond2025Pensez: Less Data, Better Reasoning -- Rethinking French LLM2025PromptCoT 2.0: Scaling Prompt Synthesis for Large Language Model Reasoning2025ScaleDiff: Scaling Difficult Problems for Advanced Mathematical Reasoning2025From Harm to Help: Turning Reasoning In-Context Demos into Assets for Reasoning LMs2025Meta-Awareness Enhances Reasoning Models: Self-Alignment Reinforcement Learning2025Skill-Targeted Adaptive Training2025A^2FM: An Adaptive Agent Foundation Model for Tool-Aware Hybrid Reasoning2025Shorter but not Worse: Frugal Reasoning via Easy Samples as Length Regularizers in Math RLVR2025Can LLMs Guide Their Own Exploration? Gradient-Guided Reinforcement Learning for LLM Reasoning2025Scaling Reasoning without Attention2025