Robotics · 2026-10-07
Commit While Futures Agree: Consequence-Aware Adaptive Action Chunking for Robot Manipulation
Yuyan Li, Yujia Wang, Yusong Huang, Junjie Yang, Yanggang Sheng, Ziyi Shi, Wenpeng Xu, Xiaoyang Zhou, Haoang Li, Hongliang Lu, Xinhu Zheng
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
CA$^3$C is an inference-time action-chunk execution framework for robot manipulation. It samples multiple candidate action chunks, uses an action-conditioned world model to imagine their consequences under shared sampling noise, and computes cross-candidate future disagreement. It then uses Bayesian change-point inference to choose a trustworthy execution horizon (“commit while imagined futures agree”) and selects the candidate closest to future consensus.
Why it matters
Action-chunking policies often execute a fixed-length prefix, but execution reliability is state-dependent, especially near contact or under distribution shift. The authors’ consequence-aware commitment adapts replanning time based on agreement among imagined futures (not action-space similarity), aiming to improve both success rate and reliability during contact-sensitive stages.
Method
- Sample K candidate action chunks from a frozen action-chunking policy at each replanning point.
- Use an action-conditioned world model to imagine future latent trajectories for each candidate while reusing the same world-model sampling noise across candidates within a replanning step.
- Estimate a trustworthy execution horizon via Bayesian change-point inference on candidate-induced future disagreement; then select the candidate closest to future consensus and execute its first h* actions.
Limitation
The authors state CA$^3$C requires a generative policy with meaningful candidate diversity and an action-sensitive world model; nearly deterministic policies and action-insensitive or inaccurate world models can cause false agreement or divergence.
Abstract (from arXiv)
Action-chunking policies predict multi-step control sequences, but a fundamental question remains: how much of a predicted action chunk should be committed before replanning? Existing systems typically execute a fixed-length prefix, implicitly assuming that the same execution horizon remains trustworthy across states. Some adaptive methods estimate this horizon from the similarity or stability of predicted actions. However, different actions may lead to the same successful outcome, whereas similar actions can produce different futures, suggesting that commitment should be determined by agreement among imagined futures rather than by similarity in action space. To this end, we propose Consequence-Aware Adaptive Action Chunking (CA$^3$C), an inference-time framework built on a simple principle: commit while imagined futures agree, and replan when they diverge. Without modifying or retraining the base policy, CA$^3$C uses an action-conditioned world model to imagine the future consequences of multiple candidate action chunks under the same sampling noise. Using these imagined consequences, we formulate execution-horizon estimation as a Bayesian change-point inference problem and select the execution candidate through future consensus. Across multiple simulation benchmarks and real-world robot manipulation tasks, CA$^3$C consistently improves diverse action-chunking policies, achieving up to a 71.8% relative reduction in failure rate over the corresponding base policies.
Related papers
- Don't Throw Away the Tail: Action Upcycling for Policy Acceleration
- Staircase Policy: Streaming Inference for World-Action Models with Large Action Chunks
- Plausible Future, Wrong Choice: Steering World-Action Models toward Intended Future
- Completion Aware Guidance for World Action Models
- LeapBot-JAC: Joint Action Chains for World Action Models
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-06 · Indexed by Pub-AI: 2026-10-07 UTC
Commit While Futures Agree: Consequence-Aware Adaptive Action Chunking for Robot Manipulation
Preprint posted: 2026-10-06 · Original record ↗
- arxiv
- 2610.07949
Explicit source versions (1)
- v1 · 2026-10-06
- Preprint available
Source date: 2026-10-06 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
Action-chunking policies predict multi-step control sequences, but a fundamental question remains: how much of a predicted action chunk should be committed before replanning? Existing systems typically execute a fixed-length prefix, implicitly assuming that the same execution horizon remains trustworthy across states. Some adaptive methods estimate this horizon from the similarity or stability of predicted actions. However, different actions may lead to the same successful outcome, whereas similar actions can produce different futures, suggesting that commitment should be determined by agreement among imagined futures rather than by similarity in action space. To this end, we propose Consequence-Aware Adaptive Action Chunking (CA$^3$C), an inference-time framework built on a simple principle: commit while imagined futures agree, and replan when they diverge. Without modifying or retraining the base policy, CA$^3$C uses an action-conditioned world model to imagine the future consequences of multiple candidate action chunks under the same sampling noise. Using these imagined consequences, we formulate execution-horizon estimation as a Bayesian change-point inference problem and select the execution candidate through future consensus. Across multiple simulation benchmarks and real-world robot manipulation tasks, CA$^3$C consistently improves diverse action-chunking policies, achieving up to a 71.8% relative reduction in failure rate over the corresponding base policies.