Multi-Armed Bandit on hard MAB instance
53.1Mean Average RewardUCB
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| UCBStrategy Category=Classic Strategies2026.04 | 53.1 | 2 | 66 | 13.68 | |
| DORABackbone=Llama-3.1 8B, Lambda (λ)=λexp2026.04 | 51.4 | 0 | 58.5 | 16.61 | |
| TSStrategy Category=Classic Strategies2026.04 | 51.2 | 0 | 57 | 17.16 | |
| GreedyStrategy Category=Classic Strategies2026.04 | 50.6 | 42 | 53 | 18.9 | |
| ε-GreedyStrategy Category=Classic Strategies2026.04 | 49 | 30 | 45.5 | 21.8 | |
| Temperature ScalingBackbone=Llama-3.1 8B, Temperature (τ)=1.52026.04 | 44 | 25 | 30.1 | 28.45 | |
| Adaptive Temperature ScalingBackbone=Llama-3.1 8B, Temperature (τ)=τexp2026.04 | 43.3 | 10 | 43.4 | 24.41 | |
| Temperature ScalingBackbone=Llama-3.1 8B, Temperature (τ)=02026.04 | 41 | 90 | 10 | 36 | |
| Temperature ScalingBackbone=Llama-3.1 8B, Temperature (τ)=0.72026.04 | 41 | 90 | 10 | 36 | |
| Temperature ScalingBackbone=Llama-3.1 8B, Temperature (τ)=12026.04 | 40.1 | 95 | 4 | 38.39 |