List your company

General Reasoning on MMLU-R, MMLU-P, and GPQA

92.3MMLU-R (multi@5 Acc)UFO-5turn
90.42890.91491.491.886May 29, 2026

Evaluation results

MethodLinks
UFO-5turnModel=Qwen2.5-7B-Instruct, Maximum budget=5 turns2026.0592.370.673.278.7
DRIFT-5turnModel=Qwen2.5-7B-Instruct, Maximum budget=5 turns2026.0591.271.272.778.4
BaseModel=Qwen2.5-7B-Instruct, Maximum budget=5 turns2026.0590.568.966.175.2
SFT-5turnModel=Qwen2.5-7B-Instruct, Maximum budget=5 turns2026.0590.569.368.176