List your company

Multi-task Knowledge Understanding on MMLU (Mean per-step regret)

0.13Mean Per-Step RegretPara
0.128480.138740.1490.15926Feb 23, 2026

Evaluation results

MethodLinks
ParaStrategy Category=Static Prompts2026.020.13
No-Rewrite (NoRw)Strategy Category=Base2026.020.134
EXP3Strategy Category=Non-Contextual2026.020.139
LinUCBStrategy Category=Contextual Linear2026.020.139
TSStrategy Category=Contextual Linear2026.020.139
DisambStrategy Category=Static Prompts2026.020.142
FTPLStrategy Category=Non-Contextual2026.020.143
TSStrategy Category=Non-Contextual2026.020.143
LinEXP3Strategy Category=Contextual Linear2026.020.144
LinUCB+KLStrategy Category=Contextual Linear2026.020.145
ϵ-FTRLStrategy Category=Non-Contextual2026.020.146
ClarifyStrategy Category=Static Prompts2026.020.15
SimplStrategy Category=Static Prompts2026.020.153
ExpandStrategy Category=Static Prompts2026.020.168
LinFTPLStrategy Category=Contextual Linear2026.020.168