List your company

Mathematical Reasoning on BRUMO25

99.2AccuracyVibeThinker-3B + CLR
13.29635.59857.980.202Feb 9, 2026Mar 2, 2026Mar 23, 2026Apr 13, 2026May 4, 2026May 25, 2026Jun 15, 2026

Evaluation results

MethodLinks
VibeThinker-3B + CLRParams=3B, Post-training optimization=CLR2026.0699.2——
VibeThinker-3BParams=3B, CLR reasoning enhancement=true2026.0699.2——
Kimi K2.5Params=1T2026.0698.3——
Gemini 3 ProParams=N/A2026.0698.3——
DeepSeek V3.2Params=671B2026.0696.7——
OpenAI o3 (high)Params=N/A2026.0695.8——
Grok 4Params=N/A2026.0695——
VibeThinker-3BParams=3B2026.0693.8——
VibeThinker-3BParams=3B2026.0693.8——
DeepConfModel=DeepSeek-8B, Compute Budget=@5122026.0292.621.7—
CoRefine TreeModel=DeepSeek-8B, Compute Budget=@5122026.0292.60.44—
MajorityModel=Qwen3-32B, Compute Budget=@5122026.0292.621.7—
DeepConfModel=DeepSeek-8B, Sample Count=@5122026.0292.6——
CoRefine TreeModel=DeepSeek-8B2026.0292.6——
MajorityModel=Qwen3-32B, Sampling Mode=Parallel, Sample Count=@5122026.0292.6——
DeepConfModel=PaCoRe-8B, Sample Count=@5122026.0292.6——
CoRefine TreeModel=PaCoRe-8B2026.0292.6——
GPT-OSS (high)Params=120B2026.0692.5——
DeepSeek R1 0528Params=671B2026.0692.5——
MajorityModel=DeepSeek-8B, Compute Budget=@5122026.029235.6—
DeepConfModel=Qwen3-32B, Compute Budget=@5122026.02921.37—
MajorityModel=DeepSeek-8B, Sampling Mode=Parallel, Sample Count=@5122026.0292——
DeepConfModel=DeepSeek-8B, Sample Count=@202026.0292——
DeepConfModel=Qwen3-32B, Sample Count=@5122026.0292——
GPT-5 (high)Params=N/A2026.0691.7——
MajorityModel=DeepSeek-8B, Sampling Mode=Parallel, Sample Count=@202026.0291.3——
MajorityModel=DeepSeek-8B, Sampling Mode=Sequential, Sample Count=@202026.0290.7——
MajorityModel=Qwen3-32B, Sampling Mode=Parallel, Sample Count=@202026.0290.7——
MajorityModel=Qwen3-32B, Sampling Mode=Sequential, Sample Count=@202026.0290.7——
DeepConfModel=Qwen3-32B, Sample Count=@202026.0290.7——
MajorityModel=PaCoRe-8B, Sampling Mode=Parallel, Sample Count=@5122026.0290.7——
MajorityModel=PaCoRe-8B, Sampling Mode=Parallel, Sample Count=@202026.0290.7——
MajorityModel=PaCoRe-8B, Sampling Mode=Sequential, Sample Count=@202026.0290.7——
DeepConfModel=PaCoRe-8B, Sample Count=@202026.0290.7——
CoRefine TreeModel=Qwen3-32B, Compute Budget=@5122026.0290.60.28—
CoRefine TreeModel=Qwen3-32B2026.0290.6——
Gemini 2.5 ProParams=N/A2026.0690——
GLM-4.5-AirParams=106B2026.0690——
CoRefineModel=DeepSeek-8B, Compute Budget=@5122026.0286.70.4—
CoRefineModel=Qwen3-32B, Compute Budget=@5122026.0286.70.18—
CoRefineModel=DeepSeek-8B2026.0286.7——
CoRefineModel=Qwen3-32B2026.0286.7——
CoRefineModel=PaCoRe-8B2026.0286.7——
Ministral-3-Reasoning-2512Params=14B2026.0686.7——
GPT-OSS-20B (high)Params=20B2026.0686.7——
Qwen3.5-4BParams=4B2026.0683.5——
Gemini 2.5 FlashParams=N/A2026.0683.3——
GPT-5 Nano (high)Params=N/A2026.0680.8——
PaCoRe-8BStrategy=Pass, Sample Count=@12026.0280.7——
Gemma-4-itParams=12B2026.0680.4——
DeepSeek-8BStrategy=Pass, Sample Count=@12026.0280——
Mimo7B-RL-0530Params=7B2026.0679.8——
OpenReasoning-NemotronParams=7B2026.0678.8——
Qwen3-32BStrategy=Pass, Sample Count=@12026.0278.7——
KnowRL-Nemotron-1.5BHint Setting=CBRS, Evaluation Protocol=mean@322026.0478.33——
KnowRL-Nemotron-1.5BHint Setting=CSS, Evaluation Protocol=mean@322026.0478.12——
Qwen3-4B-Thinking-2507Params=4B2026.0677.7——
QuestAHint Setting=CSS, Evaluation Protocol=mean@322026.0473.75——
QuestAHint Setting=CBRS, Evaluation Protocol=mean@322026.0473.23——
JustRLHint Setting=CSS, Evaluation Protocol=mean@322026.0470.67——
JustRLHint Setting=CBRS, Evaluation Protocol=mean@322026.0470.49——
KnowRL-Nemotron-1.5BHint Setting=w/o KP, Evaluation Protocol=mean@322026.0469.48——
Olmo-3-ThinkParams=7B2026.0669——
QuestAHint Setting=w/o KP, Evaluation Protocol=mean@322026.0467.5——
JustRLHint Setting=w/o KP, Evaluation Protocol=mean@322026.0466.88——
Phi4-Reasoning-PlusParams=14B2026.0666.5——
Nemotron-1.5BHint Setting=CSS, Evaluation Protocol=mean@322026.0465.03——
Nemotron-1.5BHint Setting=CBRS, Evaluation Protocol=mean@322026.0464.17——
Hunyuan-4B-InstructParams=4B2026.0662.7——
Nemotron-1.5BHint Setting=w/o KP, Evaluation Protocol=mean@322026.0460.73——
SmolLM3Params=3B2026.0649.2——
PPCVModel=Qwen3-32B, Mode=non-thinking mode2026.0243.33——
Phi-DecodingModel=Qwen3-32B, Mode=non-thinking mode2026.0236.67——
DyJRBackbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@256, alpha_JS=0.05, M=82026.0333.4——
Predictive DecodingModel=Qwen3-32B, Mode=non-thinking mode2026.0233.33——
Ex-GRPOBackbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@2562026.0332.8——
DyJR (Forward KL)Backbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@256, Divergence Type=Forward KL2026.0332.6——
DyJR (alpha_JS = 0.01)Backbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@256, alpha_JS=0.012026.0331——
DyJR (M = 16)Backbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@256, M=162026.0331——
DAPOBackbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@2562026.0330.4——
DyJR (M = 32)Backbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@256, M=322026.0330.2——
DPH-RLBackbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@2562026.0330.1——
Chain-of-ThoughtModel=Qwen3-32B, Mode=non-thinking mode2026.0230——
GRPOBackbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@2562026.0329.6——
DyJR (M = 64)Backbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@256, M=642026.0329.3——
DyJR (alpha_JS = 0.2)Backbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@256, alpha_JS=0.22026.0329——
RLEPBackbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@2562026.0328.9——
Guided DecodingModel=Qwen3-32B, Mode=non-thinking mode2026.0228.67——
Base ModelBackbone=Qwen3-4B-Base, Evaluation Protocol (Mean@k)=Mean@2562026.0316.6——
BaseModel Series=Qwen3-4B, Decoding Mode=Standard decoding2026.02——61.67
BaseModel Series=Qwen3-8B, Decoding Mode=Standard decoding2026.02——69.58
BaseModel Series=Qwen3-14B, Decoding Mode=Standard decoding2026.02——72.84
CCDModel Series=Qwen3-4B2026.02——65
CCDModel Series=Qwen3-8B2026.02——72.5
CCDModel Series=Qwen3-14B2026.02——75
MTIModel Series=Qwen3-4B2026.02——67.08
MTIModel Series=Qwen3-8B2026.02——70
MTIModel Series=Qwen3-14B2026.02——73.64