List your company

Mathematical Reasoning on GPQA Diamond (average@8 score)

45.3Average@8 ScoreDeepTool
27.82832.36436.941.436May 28, 2026

Evaluation results

MethodLinks
DeepToolBackbone=Qwen2.5-7B, Training Stage=+RL2026.0545.3
DeepToolBackbone=Qwen2.5-7B, Training Stage=SFT2026.0544.8
DeepToolBackbone=Qwen3-4B, Training Stage=+RL2026.0543.8
DeepToolBackbone=Qwen2.5-7B, Training Stage=w/o thinking2026.0543.1
DeepToolBackbone=Qwen3-4B, Training Stage=SFT2026.0537.4
DeepToolBackbone=Qwen3-4B, Training Stage=w/o thinking2026.0534.5
Search-R1Backbone=Qwen2.5-7B, Training Stage=Search-based2026.0533.3
Qwen3-4B-BaseBackbone=Qwen3-4B, Training Stage=Base2026.0532.6
Qwen2.5-7B-BaseBackbone=Qwen2.5-7B, Training Stage=Base2026.0532.4
Qwen2.5-7B-TIRBackbone=Qwen2.5-7B, Training Stage=TIR2026.0528.5