Large Discrete Policy: Advancing Explicit Behavior Modeling with Stochastic Iterative Scoring
Large Discrete Policy: Advancing Explicit Behavior Modeling with Stochastic Iterative Scoring
This research paper presents Large Discrete Policy (LDiP), a fully discrete behavior modeling framework that selects actions from a large vocabulary of physically plausible candidates. Unlike continuous generative models that rely on iterative denoising, LDiP employs stochastic iterative scoring to progressively re-score and prune candidates, enabling fine-grained ranking and exploration while maintaining an explicit decision process. The authors demonstrate that LDiP outperforms strong discrete and continuous baselines in autonomous driving and matches or exceeds continuous generative policies in robotic manipulation across end-to-end planning, closed-loop driving, and vision-language-action settings.
The paper introduces Large Discrete Policy (LDiP), a discrete behavior modeling framework that uses stochastic iterative scoring to select actions from a vocabulary of physically plausible candidates, demonstrating improved expressivity and interpretability compared to continuous generative models in autonomous driving and robotic manipulation tasks.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- code: not_reported
- weights: not_reported
- dataset: not_reported
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
tperforms strong discrete and continuous baselines in autonomous driving, and exceeds or matches continuous generative policies in robotic manipulation. These results show that discrete policies, when equipped with effective scoring mechanisms, offer an expressive, plausible, and interpretable alternative for behavior modeling. Project website: https://zhenxinli.net/LargeDiscretePolicy/.
Open source S4
Source：arXiv Robotics — research abstracts · arxiv.org