Understanding Behavior Cloning with Action Quantization

ORID 9uENnRAcSl · tags icml2026-repro paper-9uENnRAcSl

#StatusPageArtifactClaim excerpt
1VERIFIED 2/201-behavior-cloning-quantized-actions-log-lossartifactBehavior cloning with quantized actions and log-loss is proven to achieve sample…
2VERIFIED 2/202-probabilistic-incremental-input-to-state-stabiliartifactUnder Probabilistic Incremental Input-to-State Stability (P-IISS) of the dynamic…
3VERIFIED 2/203-without-smoothness-assumption-quantizer-non-smooartifactTheorem 6 shows that without a smoothness assumption on the quantizer, non-smoot…
4VERIFIED 2/204-model-based-data-augmentation-improves-horizonartifactTheorem 7 proves that model-based data augmentation improves the horizon depende…
5VERIFIED 2/205-information-theoretic-lower-bounds-theorems-estaartifactInformation-theoretic lower bounds (Theorems 8-9) establish that regret must sca…
6VERIFIED 2/206-empirically-binning-quantizers-are-shownartifactEmpirically, binning quantizers are shown to preserve policy smoothness better t…

Open logbook index · logbook.json