List your company

Mathematical Theorem Proving on MATH 500

91.2Proof-level AccuracyRaw CoT
23.28840.91958.5576.181Jun 14, 2026Jun 16, 2026Jun 18, 2026Jun 21, 2026Jun 23, 2026Jun 25, 2026Jun 28, 2026

Evaluation results

MethodLinks
Raw CoTLLM=DEEPSEEK-R1-0528-QWEN3-8B2026.0691.23,544
CLSRLLM=QWEN3-32B2026.0689.4206
Raw CoTLLM=QWEN3-32B2026.0689.3845
CLSRLLM=DEEPSEEK-R1-0528-QWEN3-8B2026.0688.11,753
Raw CoTLLM=QWEN3-8B2026.0687.2878
CLSRLLM=QWEN3-8B2026.0686.8257
SoTLLM=DEEPSEEK-R1-0528-QWEN3-8B2026.0684.11,982
SoTLLM=QWEN3-32B2026.0683.8278
CCoTLLM=DEEPSEEK-R1-0528-QWEN3-8B2026.0683.42,032
CCoTLLM=QWEN3-32B2026.0682382
CoDLLM=QWEN3-32B2026.0681.5365
SoTLLM=QWEN3-8B2026.0681.3294
CoDLLM=DEEPSEEK-R1-0528-QWEN3-8B2026.0680.51,842
CCoTLLM=QWEN3-8B2026.0678.4423
CoDLLM=QWEN3-8B2026.0674.5389
Raw CoTLLM=LLaMA3-8B2026.0651.6704
CLSRLLM=LLaMA3-8B2026.0648.2257
CCoTLLM=LLaMA3-8B2026.0645.6315
SoTLLM=LLaMA3-8B2026.0645.2289
CoDLLM=LLaMA3-8B2026.0642.2280
BASESolver=Gemini 3 Pro, Evaluation Protocol=BASE, Prover=DeepSeek-Prover-V2, K=82026.0631.4—
BASESolver=R1-Qwen3-8B, Evaluation Protocol=BASE, Prover=DeepSeek-Prover-V2, K=82026.0628.4—
BASESolver=Qwen3-8B, Evaluation Protocol=BASE, Prover=DeepSeek-Prover-V2, K=82026.0625.9—