# RePO-VLA: Recovery-Driven Policy Optimization for Vision-Language-Action Models

arXiv: https://arxiv.org/abs/2605.09410 · HTML: https://mlsanity.com/paper/2605.09410/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Failure recovery and robustness, one of 13 directions in Vision-language-action models.
- Builds on 21 of its 44 references that are in the map.
- First posted on arXiv 10 May 2026.

## Placed in

- Vision-language-action models / [Failure recovery and robustness](https://mlsanity.com/vla/failure-recovery-and-robustness/)

## Builds on (most cited first, 21 of 21)

- [Do As I Can, Not As I Say: Grounding Language in Robotic Affordances](https://mlsanity.com/paper/2204.01691/) (2022)
- [Mastering Diverse Domains through World Models](https://mlsanity.com/paper/2301.04104/) (2023)
- [Inner Monologue: Embodied Reasoning through Planning with Language Models](https://mlsanity.com/paper/2207.05608/) (2022)
- [V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning](https://mlsanity.com/paper/2506.09985/) (2025)
- [RoboTwin 2.0: A Scalable Data Generator and Benchmark with Strong Domain Randomization for Robust Bimanual Robotic Manipulation](https://mlsanity.com/paper/2506.18088/) (2025)
- [AgiBot World Colosseo: A Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systems](https://mlsanity.com/paper/2503.06669/) (2025)
- [π*0.6: a VLA That Learns From Experience](https://mlsanity.com/paper/2511.14759/) (2025)
- [REFLECT: Summarizing Robot Experiences for Failure Explanation and Correction](https://mlsanity.com/paper/2306.15724/) (2023)
- [Yell At Your Robot: Improving On-the-Fly from Language Corrections](https://mlsanity.com/paper/2403.12910/) (2024)
- [Human-in-the-Loop Imitation Learning using Remote Teleoperation](https://mlsanity.com/paper/2012.06733/) (2020)
- [GRAPE: Generalizing Robot Policy via Preference Alignment](https://mlsanity.com/paper/2411.19309/) (2024)
- [ThriftyDAgger: Budget-Aware Novelty and Risk Gating for Interactive Imitation Learning](https://mlsanity.com/paper/2109.08273/) (2021)
- [Failure Prediction at Runtime for Generative Robot Policies](https://mlsanity.com/paper/2510.09459/) (2025)
- [RePLan: Robotic Replanning with Perception and Language Models](https://mlsanity.com/paper/2401.04157/) (2024)
- [RoboFAC: A Comprehensive Framework for Robotic Failure Analysis and Correction](https://mlsanity.com/paper/2505.12224/) (2025)
- [FailSafe: Reasoning and Recovery from Failures in Vision-Language-Action Models](https://mlsanity.com/paper/2510.01642/) (2025)
- [dVLA: Diffusion Vision-Language-Action Model with Multimodal Chain-of-Thought](https://mlsanity.com/paper/2509.25681/) (2025)
- [RoboCopilot: Human-in-the-loop Interactive Imitation Learning for Robot Manipulation](https://mlsanity.com/paper/2503.07771/) (2025)
- [π_textttRL: Online RL Fine-tuning for Flow-based Vision-Language-Action Models](https://mlsanity.com/paper/2510.25889/) (2025)
- [A1: A Fully Transparent Open-Source, Adaptive and Efficient Truncated Vision-Language-Action Model](https://mlsanity.com/paper/2604.05672/) (2026)
- [RoboMD: Uncovering Robot Vulnerabilities through Semantic Potential Fields](https://mlsanity.com/paper/2412.02818/) (2024)

## Built on by (most cited first, 2 of 2)

- [Causeway: Restoring Task Accessibility for Instruction Switching in VLA Policies](https://mlsanity.com/paper/2609.30913/) (2026)
- [CAPABLE: Capability-Aware Policy Adaptation via Behavioral Latent Encoding](https://mlsanity.com/paper/2610.11971/) (2026)

## Abstract (arXiv)

Vision-Language-Action (VLA) models remain brittle in long-horizon, contact-rich manipulation because success-only imitation provides little supervision for execution drift, while failed rollouts are often discarded. We introduce RePO-VLA, a recovery-driven policy optimization framework that assigns distinct roles to success, recovery, and failure trajectories. RePO-VLA first applies Recovery-Aware Initialization (RAI), slicing recovery segments and resetting history so corrective actions depend on the current adverse state rather than the preceding failure. It then learns a Progress-Aware Semantic Value Function (PAS-VF), aligning spatiotemporal trajectory features with instructions and successful references. The resulting labels salvage useful failure prefixes via reliability decay, while low-value labels mark drift and terminal breakdowns, teaching differences among nominal, failed, and corrective actions. The data engine turns adverse states into planner-generated or human-collected corrective rollouts, teaching recovery to the success manifold. Value-Conditioned Refinement (VCR) trains the policy to prefer high-progress actions. At deployment, a fixed high value (v=1.0) biases actions toward the learned success manifold without online failure detectors or heuristic retries. We introduce FRBench, with standardized error injection and recovery-focused evaluation. Across simulated and real-world bimanual tasks, RePO-VLA improves robustness, raising adversarial success from 20% to 75% on average and up to 80% in scaled real-world trials.