Mathematical Reasoning on AIME 25 (accuracy)
94.5AccuracyAPRM
Evaluation Results
| Method | Links | |
|---|---|---|
| APRMBackbone=GPT-OSS-120B2025.11 | 94.5 | |
| APRMBackbone=GPT-OSS-20B2025.11 | 93.6 | |
| Self-ConsistencyModel=DeepSeek-V3.2, API Calls=O(N)2026.01 | 93.3 | |
| Length-based RankingModel=DeepSeek-V3.2, API Calls=O(N)2026.01 | 93.3 | |
| Self-VerificationModel=DeepSeek-V3.2, API Calls=O(NL)2026.01 | 93.3 | |
| UniCogModel=DeepSeek-V3.2, API Calls=O(N)2026.01 | 93.3 | |
| LLM-JBackbone=GPT-OSS-20B2025.11 | 92.8 | |
| APRMBackbone=Gemma-3-27B2025.11 | 91.9 | |
| LLM-JBackbone=Gemma-3-27B2025.11 | 91.6 | |
| StandardModel=DeepSeek-V3.2, API Calls=O(1)2026.01 | 91.3 | |
| ReST-MCTSBackbone=GPT-OSS-20B2025.11 | 91.3 | |
| ReST-MCTSBackbone=GPT-OSS-120B2025.11 | 91 | |
| rStar MathBackbone=GPT-OSS-20B2025.11 | 90.8 | |
| MajorityModel=PaCoRe-8B, Sampling Mode=Parallel, Sample Count=@5122026.02 | 90.7 | |
| rStar MathBackbone=GPT-OSS-120B2025.11 | 90.7 | |
| ReST-MCTSBackbone=Gemma-3-27B2025.11 | 90.6 | |
| LLM-JBackbone=GPT-OSS-120B2025.11 | 90.4 | |
| ToTBackbone=GPT-OSS-20B2025.11 | 90.4 | |
| rStar MathBackbone=Gemma-3-27B2025.11 | 89.5 | |
| ToTBackbone=GPT-OSS-120B2025.11 | 89.2 | |
| AutoPSVBackbone=GPT-OSS-120B2025.11 | 88 | |
| ToTBackbone=Gemma-3-27B2025.11 | 88 | |
| Math S.Backbone=GPT-OSS-120B2025.11 | 87.5 | |
| DeepConfModel=DeepSeek-8B, Sample Count=@5122026.02 | 87.4 | |
| DeepConfModel=DeepSeek-8B, Sample Count=@202026.02 | 87.4 | |
| DeepConfModel=PaCoRe-8B, Sample Count=@202026.02 | 87.4 | |
| CoRefine TreeModel=DeepSeek-8B2026.02 | 87.3 | |
| CoRefine TreeModel=PaCoRe-8B2026.02 | 87.3 | |
| MajorityModel=DeepSeek-8B, Sampling Mode=Parallel, Sample Count=@202026.02 | 86.7 | |
| CoRefineModel=DeepSeek-8B2026.02 | 86.7 | |
| DeepConfModel=PaCoRe-8B, Sample Count=@5122026.02 | 86.7 | |
| CoRefineModel=PaCoRe-8B2026.02 | 86.7 | |
| MajorityModel=DeepSeek-8B, Sampling Mode=Sequential, Sample Count=@202026.02 | 86 | |
| Verify SbS.Backbone=GPT-OSS-120B2025.11 | 86 | |
| CoT-SCBackbone=GPT-OSS-120B2025.11 | 85.5 | |
| AutoPSVBackbone=GPT-OSS-20B2025.11 | 85 | |
| DeepConfModel=Qwen3-32B, Sample Count=@202026.02 | 83.3 | |
| CoRefine TreeModel=Qwen3-32B2026.02 | 83.3 | |
| MajorityModel=PaCoRe-8B, Sampling Mode=Parallel, Sample Count=@202026.02 | 83.3 | |
| MajorityModel=PaCoRe-8B, Sampling Mode=Sequential, Sample Count=@202026.02 | 83.3 | |
| CoRefineModel=Qwen3-32B2026.02 | 82.7 | |
| MajorityModel=Qwen3-32B, Sampling Mode=Parallel, Sample Count=@202026.02 | 82.6 | |
| MajorityModel=DeepSeek-8B, Sampling Mode=Parallel, Sample Count=@5122026.02 | 82 | |
| MajorityModel=Qwen3-32B, Sampling Mode=Sequential, Sample Count=@202026.02 | 82 | |
| Math S.Backbone=GPT-OSS-20B2025.11 | 82 | |
| DeepConfModel=Qwen3-32B, Sample Count=@5122026.02 | 81.3 | |
| Falcon-H1R# Param.=7B2026.02 | 81.04 | |
| MajorityModel=Qwen3-32B, Sampling Mode=Parallel, Sample Count=@5122026.02 | 80.7 | |
| Verify SbS.Backbone=GPT-OSS-20B2025.11 | 80 | |
| AutoPSVBackbone=Gemma-3-27B2025.11 | 80 | |
| MiniCPM-SALA# Param.=9B2026.02 | 78.33 | |
| Math S.Backbone=Gemma-3-27B2025.11 | 78 | |
| DeepSeek-8BStrategy=Pass, Sample Count=@12026.02 | 77.4 | |
| Ministral-3-R# Param.=8B2026.02 | 75 | |
| Verify SbS.Backbone=Gemma-3-27B2025.11 | 75 | |
| DyLANMAS-Arch=Adaptive, ProVe Level=Iteration, Verifier Type=RM2026.02 | 73.33 | |
| PaCoRe-8BStrategy=Pass, Sample Count=@12026.02 | 73.3 | |
| DyLANMAS-Arch=Adaptive, ProVe Level=Agent, Verifier Type=RM2026.02 | 72.22 | |
| MiniCPM-4.1# Param.=8B2026.02 | 72.08 | |
| DyLANMAS-Arch=Adaptive, ProVe Level=Iteration, Verifier Type=Judge2026.02 | 71.11 | |
| Qwen3-32BStrategy=Pass, Sample Count=@12026.02 | 70.7 | |
| DyLANMAS-Arch=Adaptive, ProVe Level=Agent, Verifier Type=Judge2026.02 | 70 | |
| DemyAgent-4BStarting Checkpoint=Inst, Backbone Architecture Family=Qwen3-4B2026.02 | 70 | |
| CoT-SCBackbone=GPT-OSS-20B2025.11 | 69.9 | |
| CoT-SCBackbone=Gemma-3-27B2025.11 | 69.9 | |
| ELPO-4BStarting Checkpoint=Inst, Backbone Architecture Family=Qwen3-4B2026.02 | 69.4 | |
| MaASMAS-Arch=Adaptive, ProVe Level=Iteration, Verifier Type=Judge2026.02 | 68.06 | |
| DyLANMAS-Arch=Adaptive, ProVe Level=Agent, Verifier Type=PRM2026.02 | 67.78 | |
| DyLANMAS-Arch=Adaptive, ProVe Level=Iteration, Verifier Type=PRM2026.02 | 67.77 | |
| Qwen3# Param.=8B2026.02 | 66.67 | |
| APRMBackbone=Gemma-3-12B2025.11 | 65.7 | |
| DyLANMAS-Arch=Adaptive, ProVe Level=No ProVe, Verifier Type=None2026.02 | 65.55 | |
| MaASMAS-Arch=Adaptive, ProVe Level=Iteration, Verifier Type=RM2026.02 | 63.89 | |
| MaASMAS-Arch=Adaptive, ProVe Level=Iteration, Verifier Type=PRM2026.02 | 63.88 | |
| DebateMAS-Arch=Fixed, ProVe Level=Agent, Verifier Type=Judge2026.02 | 62.97 | |
| ReST-MCTSBackbone=Gemma-3-12B2025.11 | 62.6 | |
| rStar MathBackbone=Gemma-3-12B2025.11 | 62 | |
| DebateMAS-Arch=Fixed, ProVe Level=No ProVe, Verifier Type=None2026.02 | 61.87 | |
| ToTBackbone=Gemma-3-12B2025.11 | 61.4 | |
| ADASMAS-Arch=Fixed, ProVe Level=Agent, Verifier Type=RM2026.02 | 61.1 | |
| ADASMAS-Arch=Fixed, ProVe Level=Iteration, Verifier Type=Judge2026.02 | 61.1 | |
| MaASMAS-Arch=Adaptive, ProVe Level=Agent, Verifier Type=Judge2026.02 | 61.09 | |
| AFlowMAS-Arch=Fixed, ProVe Level=Iteration, Verifier Type=Judge2026.02 | 60.39 | |
| ADASMAS-Arch=Fixed, ProVe Level=Iteration, Verifier Type=RM2026.02 | 59.73 | |
| AFlowMAS-Arch=Fixed, ProVe Level=No ProVe, Verifier Type=None2026.02 | 59.72 | |
| MaASMAS-Arch=Adaptive, ProVe Level=Agent, Verifier Type=PRM2026.02 | 59.7 | |
| AFlowMAS-Arch=Fixed, ProVe Level=Agent, Verifier Type=Judge2026.02 | 58.33 | |
| ADASMAS-Arch=Fixed, ProVe Level=Agent, Verifier Type=Judge2026.02 | 58.33 | |
| MaASMAS-Arch=Adaptive, ProVe Level=No ProVe, Verifier Type=None2026.02 | 58.33 | |
| AFlowMAS-Arch=Fixed, ProVe Level=Agent, Verifier Type=RM2026.02 | 58.32 | |
| AutoPSVBackbone=Gemma-3-12B2025.11 | 58 | |
| DebateMAS-Arch=Fixed, ProVe Level=Iteration, Verifier Type=RM2026.02 | 57.78 | |
| DebateMAS-Arch=Fixed, ProVe Level=Agent, Verifier Type=PRM2026.02 | 57.12 | |
| Math S.Backbone=Gemma-3-12B2025.11 | 57 | |
| AFlowMAS-Arch=Fixed, ProVe Level=Agent, Verifier Type=PRM2026.02 | 56.94 | |
| Nemotron-Nano-v2# Param.=9B2026.02 | 56.67 | |
| LLM-JBackbone=Gemma-3-12B2025.11 | 55.4 | |
| Verify SbS.Backbone=Gemma-3-12B2025.11 | 55 | |
| DebateMAS-Arch=Fixed, ProVe Level=Iteration, Verifier Type=PRM2026.02 | 54.59 | |
| DebateMAS-Arch=Fixed, ProVe Level=Agent, Verifier Type=RM2026.02 | 54.45 |