List your company

Mathematical Reasoning on Average GSM, GSM-Hard, SVAMP, MultiArith

79.15Pass@1SFT-CoT
25.5939.49553.467.305May 25, 2026

Evaluation results

MethodLinks
SFT-CoTBackbone=Qwen3-4B2026.0579.1570.050.348
Ours-RLBackbone=Qwen3-4B2026.0572.7333.170.543
Ours-SFTBackbone=Qwen3-4B2026.0566.1934.230.335
SFT-CoTBackbone=Llama-3.2-1B2026.0556.8762.130.515
RoTBackbone=Qwen3-4B2026.0555.4320.022
SFT-w/o CoTBackbone=Qwen3-4B2026.0554.710—
Ours-RLBackbone=Llama-3.2-1B2026.0554.0725.871.021
Ours-SFTBackbone=Llama-3.2-1B2026.0550.8225.420.911
CoLaR-2Backbone=Qwen3-4B2026.0547.331.8-0.233
CoconutBackbone=Qwen3-4B2026.0531.66-3.852
CoLaR-2Backbone=Llama-3.2-1B2026.0531.429.840.126
CoconutBackbone=Llama-3.2-1B2026.0530.7460.468
SFT-w/o CoTBackbone=Llama-3.2-1B2026.0527.650—