List your company

Mathematical Reasoning on GSM Hard (Accuracy, Micro avg.)

52.69AccuracyTraceLift
3.019615.914828.8141.7052May 5, 2026

Evaluation results

MethodLinks
TraceLiftModel=Qwen2.5-14B2026.0552.6972.25
Exec-onlyModel=Qwen2.5-14B2026.0552.3971.93
TraceLiftModel=Qwen2.5-7B2026.0552.1670.19
BaseModel=Qwen2.5-14B2026.0552.1671.2
Exec-onlyModel=Qwen2.5-7B2026.0547.9265.62
BaseModel=Qwen2.5-7B2026.0530.0246.51
TraceLiftModel=Llama3.1-8B2026.051022.86
Exec-onlyModel=Llama3.1-8B2026.059.7822.19
BaseModel=Llama3.1-8B2026.054.9312.74