Contextual Bandits on Synthetic Benchmarks T=10,000, R=5
26,222.5Friedman ScoreLinearUCB
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| LinearUCBsettings=fixed2026.05 | 26,222.5 | 26,214.7 | 23,106.6 | 7,743.7 | 682.1 | 26,304.4 | 79.2 | 274 | 6 | |
| LinearTStuning=pilot run2026.05 | 15,690.6 | 26,214.7 | 10,427.3 | 6,529.8 | 524.8 | 5,260.7 | 66.9 | 301.1 | 5.125 | |
| NeuralTStuning=pilot run2026.05 | 5,697.3 | 20,311.5 | 9,673.3 | 1,376.3 | 148.2 | 728.6 | 520.7 | 198.2 | 4 | |
| RFTSsettings=fixed2026.05 | 2,504.5 | 2,562.1 | 2,688.8 | 7,579.2 | 1,058.4 | 909.6 | 781.7 | 106.7 | 4.875 | |
| XGBoostTSsettings=fixed2026.05 | 2,354.4 | 2,432.1 | 2,694.1 | 4,962.3 | 780.9 | 1,738.6 | 704.9 | 66.7 | 4.25 | |
| BFTSwarm-up=52026.05 | 680.3 | 777.1 | 835.5 | 684.4 | 367.1 | 559.3 | 570.4 | 50.9 | 2.375 | |
| PFN-TSgeometric base=2, warm-up=5, context-window truncation=none2026.05 | 233 | 243.2 | 306.3 | 568.6 | 90.2 | 136.4 | 261.4 | 51.2 | 1.375 |