List your company

Offline replay comparison on smoke-scale artifacts Abstention benchmark (offline replay)

67.5Success RateOracle
33.742.47551.2560.025Apr 30, 2026

Evaluation results

MethodLinks
Oracle2026.0467.505025-5.9
RSCB-MC2026.0462.503525-31.74
Risk-TS2026.0460037.522.5-33.7
UCB12026.04552.55017.5-35.865
Epsilon-greedy2026.0453.88.73021.3-49.818
Thompson2026.0453.72.538.820-52.315
Lexical2026.045017.5025-70.8
Static2026.045017.5025-70.8
LinUCB2026.0437.5105015-76.775
Static+Abs2026.04352.57510-33.35