Online Prediction on Online Flat length T = 100
7.9Mean RegretDeepSeek-V3
Evaluation Results
| Method | Links | |
|---|---|---|
| DeepSeek-V3Protocol=history-free note (+note)2026.05 | 7.9 | |
| DeepSeek-V3Protocol=multi-turn no-note (-note)2026.05 | 8.2 | |
| Qwen-3-14BProtocol=multi-turn no-note (-note)2026.05 | 9.7 | |
| Qwen-3-14BProtocol=history-free note (+note)2026.05 | 10.4 |