ChunkTrust: Adapting Execution Horizons for Robot Policies with Action-Expert Evidence
cs.RO
Submitted: 2026-09-30
Updated: 2026-10-10
License: http://creativecommons.org/licenses/by/4.0/
The gist: Robot foundation policies predict action chunks, but how many actions to execute before replanning depends on the current task phase.
Terminology
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/
Sources
- Qwen3-VL Technical Report
- Motus: A Unified Latent Action World Model
- GR00T N1: An Open Foundation Model for Generalist Humanoid Robots
- $\pi_0$: A Vision-Language-Action Flow Model for General Robot Control
- Dynamic Execution Commitment of Vision-Language-Action Models
- RoboTwin 2.0: A Scalable Data Generator and Benchmark with Strong Domain Randomization for Robust Bimanual Robotic Manipulation
- DREAM-Chunk: Reactive Action Chunking with Latent World Model
- GeoAAC: Geometry-Based Adaptive Action Chunking from Denoising Trajectories in VLA Policies
- StarVLA: A Lego-like Codebase for Vision-Language-Action Model Developing
- Denoising Tells When to Replan: Denoising-Variance Adaptive Chunking for Flow-Based Robot Policies
- Action ControlNet: A Lightweight Delay-Aware Adapter for Smooth Asynchronous Control in Vision-Language-Action Models
- ChainVLA: Chaining Vision-Language-Action Queries through a Unified Execution State for Long-Horizon Manipulation
- $\pi_{0.5}$: a Vision-Language-Action Model with Open-World Generalization
- Why Does Action Chunking Improve Behavioral Cloning Performance in Robotic Control?
- FASTER: Rethinking Real-Time Flow VLAs
- PACE: Phase-Aware Chunk Execution for Robot Policies with Action Chunking
- Spatial Attention: Adapting Execution Horizons for Diffusion Policies via Observation Sensitivity
- Taming Non-stationary Bandits: A Bayesian Approach
- Leave No Observation Behind: Real-time Correction for VLA Action Chunks
- RoboBrain 2.0 Technical Report
Related papers
- FMT x: An Efficient and Asymptotically Optimal Extension of the Fast Marching Tree for Dynamic Replanning
- MPCFormer: A physics-informed data-driven approach for explainable socially-aware autonomous driving
- RoboLab: A High-Fidelity Simulation Benchmark for Analysis of Task Generalist Policies
- ARC: A Reasoning Recipe for Robot Foundation Models
- HRDexDB: A 4D Dexterous Grasping Dataset Across Human and Multiple Robot Embodiments
- APT: Action Expert Pretraining Improves Instruction Generalization of Vision-Language-Action Policies