# Rewind-IL: Online Failure Detection and State Respawning for Imitation Learning

arXiv: https://arxiv.org/abs/2604.16683 · HTML: https://mlsanity.com/paper/2604.16683/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Placed in Language-conditioned policies and VLAs, one of 13 directions in Offline RL and imitation learning from data.
- Builds on 22 of its 34 references that are in the map.
- First posted on arXiv 17 Apr 2026.

## Placed in

- Offline RL and imitation learning from data / [Language-conditioned policies and VLAs](https://mlsanity.com/offlinerl/language-conditioned-policies-and-vlas/)

## Builds on (most cited first, 22 of 22)

- [Flow Matching for Generative Modeling](https://mlsanity.com/paper/2210.02747/) (2022)
- [Diffusion Policy: Visuomotor Policy Learning via Action Diffusion](https://mlsanity.com/paper/2303.04137/) (2023)
- [π0: A Vision-Language-Action Flow Model for General Robot Control](https://mlsanity.com/paper/2410.24164/) (2024)
- [Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware](https://mlsanity.com/paper/2304.13705/) (2023)
- [RoboCasa: Large-Scale Simulation of Everyday Tasks for Generalist Robots](https://mlsanity.com/paper/2406.02523/) (2024)
- [Control Barrier Functions: Theory and Applications](https://mlsanity.com/paper/1903.11199/) (2019)
- [RoboCasa365: A Large-Scale Simulation Framework for Training and Benchmarking Generalist Robots](https://mlsanity.com/paper/2603.04356/) (2026)
- [Unpacking Failure Modes of Generative Policies: Runtime Monitoring of Consistency and Progress](https://mlsanity.com/paper/2410.04640/) (2024)
- [From Foresight to Forethought: VLM-In-the-Loop Policy Steering via Latent Alignment](https://mlsanity.com/paper/2502.01828/) (2025)
- [Can We Detect Failures Without Failure Data? Uncertainty-Aware Runtime Failure Detection for Imitation Learning Policies](https://mlsanity.com/paper/2503.08558/) (2025)
- [Failure Prediction at Runtime for Generative Robot Policies](https://mlsanity.com/paper/2510.09459/) (2025)
- [Scaling Verification Can Be More Effective than Scaling Policy Learning for Vision-Language-Action Alignment](https://mlsanity.com/paper/2602.12281/) (2026)
- [CycleVLA: Proactive Self-Correcting Vision-Language-Action Models via Subtask Backtracking and Minimum Bayes Risk Decoding](https://mlsanity.com/paper/2601.02295/) (2026)
- [MoE-DP: An MoE-Enhanced Diffusion Policy for Robust Long-Horizon Robotic Manipulation with Skill Decomposition and Failure Recovery](https://mlsanity.com/paper/2511.05007/) (2025)
- [Uncertainty Comes for Free: Human-in-the-Loop Policies with Diffusion Models](https://mlsanity.com/paper/2503.01876/) (2025)
- [TriPilot-FF: Coordinated Whole-Body Teleoperation with Force Feedback](https://mlsanity.com/paper/2602.09888/) (2026)
- [KDPE: A Kernel Density Estimation Strategy for Diffusion Policy Trajectory Selection](https://mlsanity.com/paper/2508.10511/) (2025)
- [FTACT: Force Torque aware Action Chunking Transformer for Pick-and-Reorient Bottle Task](https://mlsanity.com/paper/2509.23112/) (2025)
- [DOSE3 : Diffusion-based Out-of-distribution detection on SE(3) trajectories](https://mlsanity.com/paper/2502.16725/) (2025)
- [EVE: A Generator-Verifier System for Generative Policies](https://mlsanity.com/paper/2512.21430/) (2025)
- [Reliable Robotic Task Execution in the Face of Anomalies](https://mlsanity.com/paper/2510.23121/) (2025)
- [InFeR: Informed Failure Resilience in Learned Visual Navigation Control](https://mlsanity.com/paper/2510.24680/) (2025)

## Built on by (most cited first, 9 of 9)

- [VLA-FAIL: Efficient Task Failure Detection for Finetuned Vision-Language-Action Models](https://mlsanity.com/paper/2606.21386/) (2026)
- [PATCH: Action-Chunk-Conditioned Latent Patch Innovation Monitoring for Robot Manipulation](https://mlsanity.com/paper/2606.16690/) (2026)
- [AutoIntervene: Calibrated Intervention for Action-Chunking Imitation Learning Policies](https://mlsanity.com/paper/2608.07065/) (2026)
- [ContactGuard: Pre-Contact Execution Monitoring with Action-Conditioned Latent World Models](https://mlsanity.com/paper/2608.13438/) (2026)
- [Safe Embodied AI for Long-horizon Tasks: A Cross-layer Analysis of Robotic Manipulation](https://mlsanity.com/paper/2606.05660/) (2026)
- [EgoRecovery: Acquiring Failure Recovery Ability Through Human Recovery Demonstration](https://mlsanity.com/paper/2607.19745/) (2026)
- [CARF: Contrastive Attraction-Repulsion of Failure-Guided Flow Matching](https://mlsanity.com/paper/2609.21982/) (2026)
- [Median Temporal Ensembling: Training-Free Robust Aggregation for Action-Chunked Visuomotor Policies](https://mlsanity.com/paper/2609.27167/) (2026)
- [RESCUE: Is This Alarm Worth Acting On? Counterfactual Evaluation of VLA Runtime Monitors](https://mlsanity.com/paper/2610.13516/) (2026)

## Abstract (arXiv)

Imitation learning has enabled robots to acquire complex visuomotor manipulation skills from demonstrations, but deployment failures remain a major obstacle, especially for long-horizon action-chunked policies. Once execution drifts off the demonstration manifold, these policies often continue producing locally plausible actions without recovering from the failure. Existing runtime monitors either require failure data, over-trigger under benign feature drift, or stop at failure detection without providing a recovery mechanism. We present Rewind-IL, a training-free online safeguard framework for generative action-chunked imitation policies. Rewind-IL combines a zero-shot failure detector based on Temporal Inter-chunk Discrepancy Estimate (TIDE), calibrated with split conformal prediction, with a state-respawning mechanism that returns the robot to a semantically verified safe intermediate state. Offline, a vision-language model identifies recovery checkpoints in demonstrations, and the frozen policy encoder is used to construct a compact checkpoint feature database. Online, Rewind-IL monitors self-consistency in overlapping action chunks, tracks similarity to the checkpoint library, and, upon failure, rewinds execution to the latest verified safe state before restarting inference from a clean policy state. Experiments on real-world and simulated long-horizon manipulation tasks, including transfer to flow-matching action-chunked policies, demonstrate that policy-internal consistency coupled with semantically grounded respawning offers a practical route to improved reliability in imitation learning. Supplemental materials are available at https://sjay05.github.io/rewind-il