List your company

Scientific Reasoning on GPQA Diamond (Acc, ARI, ABO, AIRW)

64AccuracyBase Model
11.79225.34638.952.454May 5, 2026

Evaluation results

MethodLinks
Base ModelBackbone=Qwen3-30B-A3B2026.05647,2036,7686,768
SxS Interleaved (RL Final)Backbone=Qwen3-30B-A3B, Training Stage=RL Final2026.0557.112,93312,8349,468
Base ModelBackbone=Qwen3-4B2026.0555.98,5218,1088,108
Standard CoT (RL Final)Backbone=Qwen3-30B-A3B, Training Stage=RL Final2026.0551.413,96913,77313,773
SxS Interleaved (RL Final)Backbone=Qwen3-4B, Training Stage=RL Final2026.0549.315,57215,7097,738
SxS Interleaved (RL Recovery)Backbone=Qwen3-30B-A3B, Training Stage=RL Recovery2026.0546.111,09011,5136,370
SxS Interleaved (RL Recovery)Backbone=Qwen3-4B, Training Stage=RL Recovery2026.0532.513,14813,3727,103
Standard CoT (SFT)Backbone=Qwen3-30B-A3B, Training Stage=SFT2026.0526.114,81714,50214,502
SxS Interleaved (SFT)Backbone=Qwen3-30B-A3B, Training Stage=SFT2026.0523.17,5248,5783,864
Standard CoT (RL Final)Backbone=Qwen3-4B, Training Stage=RL Final2026.051916,59716,33816,338
Standard CoT (SFT)Backbone=Qwen3-4B, Training Stage=SFT2026.0517.219,37919,09419,094
SxS Interleaved (SFT)Backbone=Qwen3-4B, Training Stage=SFT2026.0513.88,85410,1064,995