List your company

Truthful Question Answering on TruthfulQA (Mean per-step regret)

0.138Mean Per-Step Regretϵ-FTRL
0.135280.153640.1720.19036Feb 23, 2026

Evaluation results

MethodLinks
ϵ-FTRLStrategy Category=Non-Contextual2026.020.138
SimplStrategy Category=Static Prompts2026.020.141
LinUCB+KLStrategy Category=Contextual Linear2026.020.15
No-Rewrite (NoRw)Strategy Category=Base2026.020.151
TSStrategy Category=Non-Contextual2026.020.155
TSStrategy Category=Contextual Linear2026.020.155
ParaStrategy Category=Static Prompts2026.020.159
LinUCBStrategy Category=Contextual Linear2026.020.161
DisambStrategy Category=Static Prompts2026.020.166
FTPLStrategy Category=Non-Contextual2026.020.167
LinEXP3Strategy Category=Contextual Linear2026.020.171
EXP3Strategy Category=Non-Contextual2026.020.173
ClarifyStrategy Category=Static Prompts2026.020.18
LinFTPLStrategy Category=Contextual Linear2026.020.18
ExpandStrategy Category=Static Prompts2026.020.206