Policy learning from action-inclusive feedback on OpenML (K ≥ 3, N ≥ 70,000)
58.41Policy AccuracyCB Policy
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CB Policy2022.06 | 58.41 | — | — | |
| AI-IGLFeedback=Action-Inclusive2022.06 | 50.11 | — | 0.79 | |
| IGL (full CI)Feedback=full CI2022.06 | 11.91 | — | 0.22 | |
| Constant Action2022.06 | — | 22.57 | — |