List your company

Mathematical Reasoning on Average (GSM8k, GSM-sym, MATH, NuminaMath, AIME)

71.41AccuracySA-GRPO
20.60633.795546.98560.1745Sep 5, 2025

Evaluation results

MethodLinks
SA-GRPOBase Model=Qwen3-4B, alpha=0.22025.0971.418941.564
GRPOBase Model=Qwen3-4B2025.0969.071,030.61.165
GRPO-O1Base Model=Qwen3-4B, alpha=0.052025.0967.75608.61.592
GRPO-ERBase Model=Qwen3-4B, alpha=0.052025.0967.385791.596
PPOBase Model=Qwen3-4B2025.0967.331,3330.606
SA-PPOBase Model=Qwen3-4B2025.0966.32758.61.26
SA-GRPOBase Model=Qwen3-1.7B, alpha=0.22025.0954.136021.795
GRPOBase Model=Qwen3-1.7B2025.0953.37762.81.509
Qwen3-4BBase Model=Qwen3-4B2025.0953.037650
GRPO-O1Base Model=Qwen3-1.7B, alpha=0.052025.0952.64572.21.652
SA-PPOBase Model=Qwen3-1.7B2025.0952.49623.61.572
GRPO-ERBase Model=Qwen3-1.7B, alpha=0.052025.0952.11666.61.474
PPOBase Model=Qwen3-1.7B2025.0948.95930.40.773
SA-GRPOBase Model=Phi-3.5-mini, alpha=0.22025.0946.19368.42.137
GRPOBase Model=Phi-3.5-mini2025.0945.91448.22.003
SA-PPOBase Model=Phi-3.5-mini2025.0945.42393.81.997
Qwen3-1.7BBase Model=Qwen3-1.7B2025.0941.48825.40
GRPO-ERBase Model=Phi-3.5-mini, alpha=0.052025.0940.84374.81.368
GRPO-O1Base Model=Phi-3.5-mini, alpha=0.052025.0940.62392.81.316
PPOBase Model=Phi-3.5-mini2025.0940.247990.784
Phi-3.5-miniBase Model=Phi-3.5-mini2025.0935.15850.40
SA-GRPOBase Model=Gemma3-1B, alpha=0.22025.0932.241,017.42.218
GRPOBase Model=Gemma3-1B2025.0931.781,8661.343
GRPO-ERBase Model=Gemma3-1B, alpha=0.052025.0931.291,236.61.808
GRPO-O1Base Model=Gemma3-1B, alpha=0.052025.0931.231,529.61.528
PPOBase Model=Gemma3-1B2025.0929.691,572.61.146
SA-PPOBase Model=Gemma3-1B2025.0929.11,1701.383
Gemma3-1BBase Model=Gemma3-1B2025.0922.561,0970