Robot policies adapt action lengths based on task phase evidence
ChunkTrust: Adapting Execution Horizons for Robot Policies with Action-Expert Evidence
Robotics
Summary
Robots often decide how many actions to carry out before planning again, but this number can depend on what stage the robot is in during a task. The paper presents ChunkTrust, a method that figures out the best number of actions to execute by looking at evidence from previous expert actions instead of using a fixed setting. It uses two components: one that analyzes stability within predicted action chunks and another that checks continuity with past actions. Testing shows this approach improves robot task success rates in both simulation and real-world household tasks.
What this means in practice
- •For robotics engineers: Improve robot task success by adapting action execution length dynamically based on task phase evidence.
- •For automation system developers: Integrate adaptive horizon selection to enhance performance of automated systems working in varied physical environments.
Authors
Fanding Huang, Jingyan Jiang, Shifeng Bao, Mingkang Pu, Shiwei Li, Jing Xu, Shijia Xu, Guanbo Huang, Chenghao Gu, Yuzhi Huang, Chenxin Li, Faisal Nadeem Khan, Huan Yang, Yan Wang, Cheng Chi, Zhi Wang
Abstract
Robot foundation policies predict action chunks, but how many actions to execute before replanning depends on the current task phase. We introduce ChunkTrust, which treats the execution horizon as a latent variable inferred from action-expert evidence rather than a fixed hyperparameter. Its training-free Action-aware Horizon Selector (AHS) combines intra-chunk spectral stability of generation traces with inter-chunk continuity between executed history and predicted actions. An online Beta posterior with kernel forgetting tracks horizon preferences across replans. A lightweight Query-based Horizon Adapter (QHA) optionally learns a context-conditioned dense prior from complementary evidence, fused with current evidence and episode-local Beta memory while the base policy remains frozen. Across RoboTwin2.0 and RoboCasa GR1 Tabletop, AHS improves overall task-averaged success for each evaluated base-policy configuration, including gains of +6.80 percentage points on $π_{0.5}$ over all 50 RoboTwin2.0 tasks and +9.67 percentage points on Qwen3GR00T in RoboCasa. AHS+QHA raises the gain over Base to +9.44 percentage points on the eight-task $π_{0.5}$ evaluation. On four real-world household tasks, AHS improves the equal-task mean normalized process score from 50.4% to 57.5%. Ablations examine the contributions of both evidence terms, temporal memory, and the learned prior. Project page is https://hf618.github.io/ChunkTrust.github.io/