List your company

Mathematical Reasoning on MATH (Topic Accuracy Breakdown)

48.6Overall AccuracyStoSignSGD
47.393647.706848.0248.3332Apr 16, 2026

Evaluation results

MethodLinks
StoSignSGDModel=Qwen2.5-7B, Fine-tuning Method=LoRA, Evaluation Protocol=Zero-shot2026.0448.670.8563.3856.6740.5133.428.7921.79
AdamWModel=Qwen2.5-7B, Fine-tuning Method=LoRA, Evaluation Protocol=Zero-shot2026.0448.0869.8463.1554.2639.6634.4528.7921.79
SignSGDModel=Qwen2.5-7B, Fine-tuning Method=LoRA, Evaluation Protocol=Zero-shot2026.0447.4469.1761.7756.1136.534.6627.3522.71