List your company
ResearchFormerly WizWand

Multi-Armed Bandit on Text-based Game

0.531Average RewardUCB
0.387480.424740.4620.49926Apr 19, 2026
Updated 6mo ago

Evaluation results

MethodLinks
UCBStrategy Category=Classic Strategies, Model=Meta Llama 3.1 8B2026.040.53126613.680
DORA (λ-sched)Strategy Category=Learned Policies, Mode=λ-sched, Model=Meta Llama 3.1 8B2026.040.514058.516.610
TSStrategy Category=Classic Strategies, Model=Meta Llama 3.1 8B2026.040.51205717.160
GreedyStrategy Category=Classic Strategies, Model=Meta Llama 3.1 8B2026.040.506425318.90
ε-GreedyStrategy Category=Classic Strategies, Model=Meta Llama 3.1 8B2026.040.493045.521.80
DORA (auto)Strategy Category=Learned Policies, Mode=auto, Model=Meta Llama 3.1 8B2026.040.4626035.625.770
Temperature SamplingStrategy Category=Temperature (τ), τ=1.5, Model=Meta Llama 3.1 8B2026.040.442530.128.456.15
τ PolicyStrategy Category=Temperature (τ), Model=Meta Llama 3.1 8B2026.040.4331043.424.4121.2
Temperature SamplingStrategy Category=Temperature (τ), τ=0, Model=Meta Llama 3.1 8B2026.040.419010360
Temperature SamplingStrategy Category=Temperature (τ), τ=0.3, Model=Meta Llama 3.1 8B2026.040.419010360
Temperature SamplingStrategy Category=Temperature (τ), τ=0.7, Model=Meta Llama 3.1 8B2026.040.419010360
Temperature SamplingStrategy Category=Temperature (τ), τ=1, Model=Meta Llama 3.1 8B2026.040.40195438.390
Temperature SamplingStrategy Category=Temperature (τ), τ=2, Model=Meta Llama 3.1 8B2026.040.393044.225.738.35