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

Agentic Task Solving on DataBench

92.7Pass@3

A³

Updated 23d ago

Evaluation Results

Method	Links
A³ 2026.05		92.7	96
GSPO 2026.05		89.9	94.7
A³ 2026.05		88.4	93.3
RetroAgent 2026.05		85.7	89.3
ReACT 2026.05		79.7	82.7
LATS 2026.05		77.6	85.3
GiGPO 2026.05		77.2	84
rStar 2026.05		72.5	80
HGPO 2026.05		70.9	81.3