Driving · 2026-10-09
RESETTLE: Robotic Recovery through Disagreement-Triggered Retrieval and Efficient Corrective Control
Yuxin Chen, Senqiao Yang, Zixuan Wang, Jinhui Ye, Changsheng Lu, Pengguang Chen, Shu Liu, Zhuotao Tian, Jiaya Jia
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
RESETTLE is a model-agnostic recovery framework for frozen robot policies. It generates two action proposals from independent noise under identical conditioning and triggers recovery when their temporally weighted disagreement persists. It retrieves a same-task demo reference using an adapted V-JEPA encoder and executes one corrective action via a state-servo prior plus guarded visual residual, then returns to the base policy.
Why it matters
The authors aim to reduce recovery latency at the action-execution interface without repeated vision-language reasoning or online trajectory optimization. They show improved task success for multiple frozen VLA/WAM base policies in simulation, better performance on real-world manipulation tasks, and much lower monitoring-and-recovery latency than VoLoAgent in QwenPI-based comparisons.
Method
- Generate two action chunks from independent noise under identical observations/task conditioning; compute temporally weighted disagreement and trigger only after two consecutive threshold exceedances.
- When triggered, retrieve a same-task demonstration reference using an adapted V-JEPA-2 encoder (cosine distance in representation space).
- Execute one corrective native control action using a state-servo prior toward the reference state plus a guarded visual residual; then return control to the frozen base policy (no online trajectory optimization or additional LLM/VLM calls).
Limitation
Disagreement-based monitoring does not distinguish correct, stable behavior from stable but incorrect behavior, and triggered intervention does not guarantee completion within the evaluation budget.
Abstract (from arXiv)
Reliable robotic manipulation requires timely intervention to correct emerging deviations and restore progress after execution errors. However, recovery methods based on repeated vision-language reasoning or iterative online optimization can incur substantial latency, delaying intervention. To address these challenges, we introduce RESETTLE(Robotic rEcovery through diSagrEement-Triggered reTrievaL and Efficient Corrective Control), a model-agnostic framework that provides computationally efficient recovery at the action-execution interface of frozen robot policies. RESETTLE triggers recovery when two action proposals independently sampled under identical conditioning persistently disagree. It retrieves a same-task demonstration reference using an adapted V-JEPA encoder and combines a state-servo prior with a guarded visual residual to execute one corrective action without online trajectory optimization or additional vision-language reasoning, then returns control to the base policy. Across six base policies in simulation, RESETTLE achieves up to 8.70%, 6.28%, and 6.83% absolute success-rate gains on LIBERO-Plus, Meta-World, and RoboCasa Tabletop, respectively, with further improvements on four real-world tasks using two policies. In QwenPI-based comparisons, its monitoring-and-recovery computation latency is 74.04%--93.57% lower than VoLoAgent's monitoring-and-planning latency for grasp and place tool calls. It also raises Harness VLA's LIBERO-Pro Swap success from 42% to 50%, demonstrating compatibility with high-level agentic planning. Code available at: https://github.com/JIA-Lab-research/RESETTLE
Related papers
- ForkCast: Predictive Recovery for Vision-Language-Action Policies via Counterfactual Futures
- Don't Throw Away the Tail: Action Upcycling for Policy Acceleration
- Future Anchored Verification and Online Recovery for World Action Models
- Commit While Futures Agree: Consequence-Aware Adaptive Action Chunking for Robot Manipulation
- RoboHarn-Evo: Evolving Hierarchical Physical Knowledge for Self-Improving Robotic Manipulation
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-08 · Indexed by Pub-AI: 2026-10-09 UTC
RESETTLE: Robotic Recovery through Disagreement-Triggered Retrieval and Efficient Corrective Control
Preprint posted: 2026-10-08 · Original record ↗
- arxiv
- 2610.12185
Explicit source versions (1)
- v1 · 2026-10-08
- Preprint available
Source date: 2026-10-08 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
Reliable robotic manipulation requires timely intervention to correct emerging deviations and restore progress after execution errors. However, recovery methods based on repeated vision-language reasoning or iterative online optimization can incur substantial latency, delaying intervention. To address these challenges, we introduce RESETTLE(Robotic rEcovery through diSagrEement-Triggered reTrievaL and Efficient Corrective Control), a model-agnostic framework that provides computationally efficient recovery at the action-execution interface of frozen robot policies. RESETTLE triggers recovery when two action proposals independently sampled under identical conditioning persistently disagree. It retrieves a same-task demonstration reference using an adapted V-JEPA encoder and combines a state-servo prior with a guarded visual residual to execute one corrective action without online trajectory optimization or additional vision-language reasoning, then returns control to the base policy. Across six base policies in simulation, RESETTLE achieves up to 8.70%, 6.28%, and 6.83% absolute success-rate gains on LIBERO-Plus, Meta-World, and RoboCasa Tabletop, respectively, with further improvements on four real-world tasks using two policies. In QwenPI-based comparisons, its monitoring-and-recovery computation latency is 74.04%--93.57% lower than VoLoAgent's monitoring-and-planning latency for grasp and place tool calls. It also raises Harness VLA's LIBERO-Pro Swap success from 42% to 50%, demonstrating compatibility with high-level agentic planning. Code available at: https://github.com/JIA-Lab-research/RESETTLE
