List your company

Mathematical Reasoning on DeepMath 103K subset

65.9AccuracyTTVS
43.22849.1145560.886Apr 9, 2026

Evaluation results

MethodLinks
TTVSBackbone=Qwen3-4B2026.0465.9
TTRLBackbone=Qwen3-4B2026.0462.8
InitBackbone=Qwen3-4B2026.0453.2
TTVSBackbone=LLaMA-3.2-3B-Instruct2026.0449
TTRLBackbone=LLaMA-3.2-3B-Instruct2026.0446.6
InitBackbone=LLaMA-3.2-3B-Instruct2026.0444.1