ResearchBenchmarksMathematical Reasoning on CodeI/OFollow5.6Pass@1TAC2.483.294.14.91Jun 23, 2026Evaluation ResultsMethodMethodLinksPass@1TACBackbone=Qwen3-4B-Base...Backbone=Qwen3-4B-Base, Seeds=12026.065.6SECBackbone=Qwen3-4B-Base...Backbone=Qwen3-4B-Base, Seeds=12026.065.2RandomBackbone=Qwen3-0.6B-Ba...Backbone=Qwen3-0.6B-Base, Seeds=12026.065.1RandomBackbone=Qwen3-4B-Base...Backbone=Qwen3-4B-Base, Seeds=12026.064.2SECBackbone=Qwen3-0.6B-Ba...Backbone=Qwen3-0.6B-Base, Seeds=12026.062.8TACBackbone=Qwen3-0.6B-Ba...Backbone=Qwen3-0.6B-Base, Seeds=12026.062.6