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

Scientific Reasoning on GPQA (Accuracy, Gain, Speedup)

27Accuracy

Full-step decoding

25.12825.61426.126.586Aug 27, 2025
Updated 1mo ago

Evaluation Results

MethodLinks
2025.08
27--
2025.08
26.6-0.41.66
2025.08
25.70.51.82
2025.08
25.2--