List your company

Long-Term Credit Assignment on Key-to-Door (test)

99.9Win RateConditional Attribute Transformers
-2.33224.20950.7577.291May 13, 2026

Evaluation results

MethodLinks
Conditional Attribute TransformersDecoding strategy=greedy2026.0599.9
Percentile Behavioral CloningTraining condition=trained on only winning examples2026.0595.1
Decision Transformers2026.0594.6
Conservative Q-Learning2026.0513.3
Random Policy2026.053.1
Behavioral Cloning2026.051.6