Physical Commonsense Reasoning on PIQA (Mean per-step regret)
0.152Mean Per-Step RegretLinFTPL
Evaluation Results
| Method | Links | |
|---|---|---|
| LinFTPLStrategy Category=Contextual Linear2026.02 | 0.152 | |
| SimplStrategy Category=Static Prompts2026.02 | 0.161 | |
| No-Rewrite (NoRw)Strategy Category=Base2026.02 | 0.172 | |
| LinEXP3Strategy Category=Contextual Linear2026.02 | 0.173 | |
| ParaStrategy Category=Static Prompts2026.02 | 0.174 | |
| TSStrategy Category=Non-Contextual2026.02 | 0.174 | |
| TSStrategy Category=Contextual Linear2026.02 | 0.186 | |
| ϵ-FTRLStrategy Category=Non-Contextual2026.02 | 0.192 | |
| LinUCB+KLStrategy Category=Contextual Linear2026.02 | 0.193 | |
| LinUCBStrategy Category=Contextual Linear2026.02 | 0.197 | |
| EXP3Strategy Category=Non-Contextual2026.02 | 0.213 | |
| ClarifyStrategy Category=Static Prompts2026.02 | 0.236 | |
| DisambStrategy Category=Static Prompts2026.02 | 0.252 | |
| FTPLStrategy Category=Non-Contextual2026.02 | 0.259 | |
| ExpandStrategy Category=Static Prompts2026.02 | 0.34 |