AI Paper Atlas

Research map / Benchmarks / AIME25

AIME25 leaderboard

82 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on AIME25. Best accuracy: 96 (HyPER, ICML 2026).

Numbers are as each paper reports them, read by AI from its results tables and not checked by hand. Papers can differ in setting (training data, model size, extra data); the topic column shows what each paper is about. Open the paper before you compare or quote a number.

AIME25 · accuracy

56 papers, each with its best reported number for its own method; higher is better.

#accuracy ↑MethodPaperTopic
196HyPER
Qwen3-Next-80B-A3B-Thinking
HyPER: Bridging Exploration and Exploitation for Scalable LLM Reasoning with Hypothesis Path Expansion and Reduction
ICML 2026 · Table 1
Reasoning and chain of thought
294.7Think@n
OSS-120B-medium
Think Deep, Not Just Long: Measuring LLM Reasoning Effort via Deep-Thinking Tokens
ICML 2026 · Table 2
Reasoning and chain of thought
394.4GSPO + IBPO(mask)
Qwen3-Next-80B-A3B-Thinking
Reducing Credit Assignment Variance via Counterfactual Reasoning Paths
NeurIPS 2026 · Table 1
Reasoning and chain of thought
486.7Hermes@1
o3-mini
HERMES: Towards Efficient and Verifiable Mathematical Reasoning in LLMs
ICML 2026 · Table 1
Reasoning and chain of thought
586.67Q-Value steering
Qwen3-4B-Thinking
Modeling Hierarchical Thinking in Large Reasoning Models
ICML 2026 · Table 1
Reasoning and chain of thought
686.67RTPurbo w/ top-p
Qwen3-30B-A3B-Think
Full Attention Strikes Back: Transferring Full Attention into Sparse within Hundred Training Steps
NeurIPS 2026 · Table 3
Attention and transformers
781.5ToCoRL
Qwen3-30B-A3B-2507-Thinking
Revealing Behavioral Plasticity in Large Language Models: A Token-Conditional Perspective
ICML 2026 · Table 2
Reasoning and chain of thought
880SxS Interleaved (RL Final)
Qwen3-4B
When to Think, When to Speak: Learning Disclosure Policies for LLM Reasoning
ICML 2026 · Table 1
Reasoning and chain of thought
978.13CUSUM early exit
Qwen3-4B-Thinking-2507
Unveiling the Entropy Dynamics of Chain-of-Thought Reasoning
ICML 2026 · Table 2
Reasoning and chain of thought
1074.67Selective Strategy Retrieval (SSR)
Qwen3-14B
Strategy Executability in Mathematical Reasoning: Leveraging Human–Model Differences for Effective Guidance
ICML 2026 · Table 1
Reasoning and chain of thought
46 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

AIME25 · pass@1

26 papers, each with its best reported number for its own method; higher is better.

#pass@1 ↑MethodPaperTopic
180HICRA
Llama-3.1-8B-Instruct
Emergent Hierarchical Reasoning in LLMs through Reinforcement Learning
ICLR 2026 · Table 1
Reasoning and chain of thought
273.6MetaStone-S1-32B-high
QwQ-32B
Test-Time Scaling with Reflective Generative Model
ICLR 2026 · Table 1
Reasoning and chain of thought
373.5CoNL
Qwen3-8B
Conversation for Non-verifiable Learning: Self-Evolving Large Language Models through Meta-Evaluation
ICML 2026 · Table 3
Agents, code and math
473.2CRD-30B-A3B
Qwen3-30B-A3B
Collaborative Reasoning Distillation via Cross-Feedback and Coherent Curation
NeurIPS 2026 · Table 1
Reasoning and chain of thought
570.8GFPO adaptive difficulty
Phi-4-reasoning (14B)
Sample More to Think Less: Group Filtered Policy Optimization for Concise Reasoning
ICLR 2026 · Table 1
Reasoning and chain of thought
667.4Adaptive control
QwQ-32B
Controlling Thinking Speed in Reasoning Models
NeurIPS 2025 · Adaptive control results table
Reasoning and chain of thought
766.7DyCon (Ours)
QwQ-32B
DyCon: Dynamic Reasoning Control via Evolving Difficulty Modeling
ICML 2026 · Table 1
Reasoning and chain of thought
863.96ProRL + GooseReason-0.7M (+270 steps)
Qwen3-4B-Instruct
Golden Goose: A Simple Trick to Synthesize Unlimited RLVR Tasks from Unverifiable Internet Text
ICML 2026 · Table 1
Reasoning and chain of thought
963XRPO
Llama-3.2-3B
XRPO: Pushing the Limits of GRPO with Targeted Exploration and Exploitation
ICML 2026 · Table 2
RL from feedback
1061.3PrefixRL
Llama-3.1-8B
Reuse your FLOPs: Scaling RL on Hard Problems by Conditioning on Very Off-Policy Prefixes
ICML 2026 · text
RL from feedback
16 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

Not affiliated with any conference, journal or the benchmark's authors. Corrections: contact@latenticlabs.com.