List your company

Reasoning on Comp-Math

76.16AccuracyQwen3-4B
56.264861.429966.59571.7601Feb 1, 2026

Evaluation results

MethodLinks
Qwen3-4BLoss function=base+aux2026.0276.16
Qwen3-4BLoss function=base2026.0275.2
Qwen3-8BLoss function=base+aux2026.0275.09
Qwen3-8BLoss function=base2026.0273.65
LLaMA3.1-8BLoss function=base+aux2026.0258.52
LLaMA3.1-8BLoss function=base2026.0257.03