Share your thoughts, 1 month free Claude Pro on usSee more

Graduate-level Q&A on GPQA 0-shot

27.5GPQA Accuracy (0-shot)

AR SFT

Updated 9d ago

Evaluation Results

Method	Links
AR SFT 2026.04		27.5
MARS-7B 2026.04		26.6
MARS-0.5B 2026.04		19.4
MARS-0.5B w/o SFT loss 2026.04		19
AR SFT (5 ep) 2026.04		17.9
Block Diffusion 2026.04		17.9
AR SFT (10 ep) 2026.04		14.1
Base (no fine-tune) 2026.04		13.6
Base (no fine-tune) 2026.04		11.6