List your company

Commonsense Reasoning on Winogrande (ACC)

80.71AccuracyLoPT-GRPO
46.847655.638864.4373.2212May 5, 2026May 6, 2026

Evaluation results

MethodLinks
LoPT-GRPOBackbone=Qwen2.5-32B2026.0580.71
BaseBackbone=Qwen2.5-32B2026.0580.35
E2E-GRPOBackbone=Qwen2.5-32B2026.0580.27
MUDDFORMEREvaluation Protocol=Zero-shot, Number of Parameters=760M, Training Tokens=20B2026.0555.64
TRANSFORMEREvaluation Protocol=Zero-shot, Number of Parameters=1.3B, Training Tokens=30B2026.0554.85
SATFORMEREvaluation Protocol=Zero-shot, Number of Parameters=1.3B, Training Tokens=30B2026.0554.3
HYPERCONNECTIONSEvaluation Protocol=Zero-shot, Number of Parameters=760M, Training Tokens=20B2026.0552.95
RESFORMEREvaluation Protocol=Zero-shot, Number of Parameters=1.3B, Training Tokens=30B2026.0552.88
RESFORMEREvaluation Protocol=Zero-shot, Number of Parameters=760M, Training Tokens=20B2026.0552.64
MUDDFORMEREvaluation Protocol=Zero-shot, Number of Parameters=130M, Training Tokens=5B2026.0552.49
DENSEFORMEREvaluation Protocol=Zero-shot, Number of Parameters=760M, Training Tokens=20B2026.0552.17
TRANSFORMEREvaluation Protocol=Zero-shot, Number of Parameters=130M, Training Tokens=5B2026.0551.85
RESFORMEREvaluation Protocol=Zero-shot, Number of Parameters=340M, Training Tokens=10B2026.0551.46
RESFORMEREvaluation Protocol=Zero-shot, Number of Parameters=130M, Training Tokens=5B2026.0551.3
MUDDFORMEREvaluation Protocol=Zero-shot, Number of Parameters=340M, Training Tokens=10B2026.0551.07
SATFORMEREvaluation Protocol=Zero-shot, Number of Parameters=340M, Training Tokens=10B2026.0551.07
HYPERCONNECTIONSEvaluation Protocol=Zero-shot, Number of Parameters=340M, Training Tokens=10B2026.0551.06
HYPERCONNECTIONSEvaluation Protocol=Zero-shot, Number of Parameters=130M, Training Tokens=5B2026.0550.98
SATFORMEREvaluation Protocol=Zero-shot, Number of Parameters=130M, Training Tokens=5B2026.0550.91
TRANSFORMEREvaluation Protocol=Zero-shot, Number of Parameters=340M, Training Tokens=10B2026.0550.83
TRANSFORMEREvaluation Protocol=Zero-shot, Number of Parameters=760M, Training Tokens=20B2026.0550.75
SATFORMEREvaluation Protocol=Zero-shot, Number of Parameters=760M, Training Tokens=20B2026.0550.67
DENSEFORMEREvaluation Protocol=Zero-shot, Number of Parameters=130M, Training Tokens=5B2026.0550.43
DENSEFORMEREvaluation Protocol=Zero-shot, Number of Parameters=340M, Training Tokens=10B2026.0548.15