mlsanityData to 7 Oct 2026

Failure recovery and robustness

109 papers5 this weekMedium confidence in this grouping: mostly one topic, with blurry edgesMedium confidence in this grouping: mostly one topic, with blurry edges · About

Keywords: failure, vla, manipulation, vision-language-action, success

Key facts. Failure recovery and robustness: 109 papers on arXiv since 2025 (data through 7 Oct 2026). 26 of its 109 mapped papers (24%) cite “LIBERO-Plus: In-depth Robustness Analysis of Vision-Language-Action Models” (2025, arXiv 2510.13626). 20 of its 109 mapped papers (18%) cite “SAFE: Multitask Failure Detection for Vision-Language-Action Models” (2025, arXiv 2506.09937). 36 new papers 17 Sep–7 Oct.

What’s new

17 Sep – 7 Oct 2026

36 new papers, 17 Sep – 7 Oct; 5 this week.

  1. This week · 5
  2. Sparse Feature Policy Unlearning Mitigates State Hallucination in Vision-Language-Action Models also fits: Reasoning and chain-of-thought in VLA
  3. Beyond Reconstruction: What Matters in Action Tokenization for Robot Policies? also fits: Efficient VLA inference and adaptation
  4. CARE: Certifying Acceleration for Vision-Language-Action Inference also fits: Efficient VLA inference and adaptation
  5. Adapting Vision-Language-Action Models to Unknown Visual Disruptions During Execution also fits: Reinforcement learning and post-training
  6. When to Switch: Reliable Action-Chunk Extension for Vision-Language-Action Models also fits: Efficient VLA inference and adaptation
  7. Earlier · 31
  8. A Safe Action Is Not Enough: Feasible-Future Decoding for Vision-Language-Action Policies also fits: Safety and adversarial attacks
  9. Does a Learned Corrector Beat a Simple Retreat? Evidence from a Frozen VLA also fits: Reinforcement learning and post-training
  10. SocialVLA: A Social Perception Gateway for Human-Reaction-Based Failure Detection and Recovery in VLA Manipulation
  11. Is Success All You Need? Investigating the Impact of Input Perturbations on VLA Behaviour in Tabletop Manipulation Tasks
  12. MIKASA-Robo-VLA: Benchmarking Memory in VLA Models for Long-Horizon Manipulation also fits: Reasoning and chain-of-thought in VLA
All 36
  1. When Instructions Retrieve Trajectories: Diagnosing and Mitigating Generalization Failures in VLA Models also fits: VLA architectures and generalization
  2. Learning from Runtime Feedback through Failure-Bank Self-Evolution for Vision-Language-Action Models
  3. Blackout vs. Freeze: Analyzing Physical Failure Modes of VLAs under Camera Faults also fits: Safety and adversarial attacks
  4. Explore, Execute, Evolve: A Skill Acquisition and Reuse Loop for Embodied Agents also fits: Reinforcement learning and post-training
  5. Faster and Better? Benchmark Bugs and Design Limitations Distort the Evaluation of Vision-Language-Action Acceleration
  6. Taming VLAs under Robot Execution Errors: Self-Compensation and Stress Testing also fits: Reinforcement learning and post-training
  7. Self-Evolving Coding Agents: From Digital Programs to Physical-World Intelligence also fits: World model-based VLA
  8. Do Not Cut When Uncertain: Rejectable and Calibrated Decision Heads for VLA Policies in Robotic Harvesting also fits: Reinforcement learning and post-training
  9. Learning to Act under Visual Interruptions with Vision-Language-Action Models
  10. Natural State-Prediction Accuracy can Hide Weak Controlled Responsiveness in VLA Readouts
  11. Gaze Prompts: Temporally Dense Human Attention for Vision-Language-Action Fine-Tuning also fits: Reasoning and chain-of-thought in VLA
  12. FailPatch: Failure Residual Patching for Vision-Language-Action Models
  13. SEES: A Self-Evolving Embodied System via Failure-Guided VLA Policy Adaptation also fits: Reinforcement learning and post-training
  14. ConflictVLA-Bench: Benchmarking Behavioral Responses of Vision-Language-Action Models to Premise Conflicts
  15. Kintsugi-VLA: Turning Failed Robot Rollouts into Recovery Data through Interventional Recoverability
  16. Causeway: Restoring Task Accessibility for Instruction Switching in VLA Policies also fits: Reasoning and chain-of-thought in VLA
  17. AdaHVLA: Adaptive Harnesses for Long-Horizon Vision-Language-Action Execution also fits: Reasoning and chain-of-thought in VLA
  18. Imperfection for Precision: Upcycling Imperfect Data for High-Precision Robotic Manipulation
  19. IndustrialVLA-Bench: A Traceable Multi-Axis Evaluation of Open Robot Policy Models also fits: Reinforcement learning and post-training
  20. CARE: Experience-Guided Atomic Corrective Execution for Vision-Language-Action Policies
  21. ReVeal: A Reconstruction-Aware Real-to-Sim Framework for VLA Policy Evaluation also fits: 3D and spatial-aware VLA
  22. Beyond Appearance Shifts: Task-Semantic Action Calibration for VLA Models also fits: VLA architectures and generalization
  23. CommitFlow: Semantic Commitment Verification and Local Correction for Long-Horizon Robot Manipulation VLA Execution
  24. Outcome-Conditioned End-Effector Geometry Across Vision-Language-Action Policies also fits: Reasoning and chain-of-thought in VLA
  25. FAN: Foresight Action Normalization for Continual Adaptation of Vision-Language-Action Models also fits: Human-video and latent-action pretraining
  26. VLA-Scope: Shift-Aware Failure Prediction for Vision-Language-Action Models

Foundations

Most cited by papers here

  1. LIBERO-Plus: In-depth Robustness Analysis of Vision-Language-Action Models2025cited by 26 here
  2. SAFE: Multitask Failure Detection for Vision-Language-Action Models2025cited by 20 here
  3. LIBERO-PRO: Towards Robust and Fair Evaluation of Vision-Language-Action Models Beyond Memorization2025cited by 18 here
  4. FailSafe: Reasoning and Recovery from Failures in Vision-Language-Action Models2025cited by 11 here
  5. RoboFAC: A Comprehensive Framework for Robotic Failure Analysis and Correction2025cited by 7 here

Lineage

The papers this direction builds on, oldest first.

Each paper on the line cites the one before.

  1. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 75 here
  2. 2024Octo: An Open-Source Generalist Robot Policycited by 52 here
  3. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 96 here
  4. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 80 here
  5. 2025FAST: Efficient Action Tokenization for Vision-Language-Action Modelscited by 26 here
  6. 2025SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Modelcited by 24 here
  7. 2025NORA: A Small Open-Sourced Generalist Vision Language Action Model for Embodied Taskscited by 10 here
  8. 2025LIBERO-Plus: In-depth Robustness Analysis of Vision-Language-Action Modelscited by 26 here
16 more it builds on, not in this chain
  1. 2024A Survey on Vision-Language-Action Models for Embodied AIcited by 12 here
  2. 2024VLATest: Testing and Evaluating Vision-Language-Action Models for Robotic Manipulationcited by 10 here
  3. 2024CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulationcited by 16 here
  4. 2024TraceVLA: Visual Trace Prompting Enhances Spatial-Temporal Awareness for Generalist Robotic Policiescited by 13 here
  5. 2025Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Successcited by 46 here
  6. 2025CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Modelscited by 10 here
  7. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 72 here
  8. 2025UniVLA: Learning to Act Anywhere with Task-centric Latent Actionscited by 16 here
  9. 2025SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Roboticscited by 23 here
  10. 2025SAFE: Multitask Failure Detection for Vision-Language-Action Modelscited by 20 here
  11. 2025ThinkAct: Vision-Language-Action Reasoning via Reinforced Visual Latent Planningcited by 9 here
  12. 2025VLA-Adapter: An Effective Paradigm for Tiny-Scale Vision-Language-Action Modelcited by 9 here
  13. 2025FailSafe: Reasoning and Recovery from Failures in Vision-Language-Action Modelscited by 11 here
  14. 2025LIBERO-PRO: Towards Robust and Fair Evaluation of Vision-Language-Action Models Beyond Memorizationcited by 18 here
  15. 2025X-VLA: Soft-Prompted Transformer as Scalable Cross-Embodiment Vision-Language-Action Modelcited by 11 here
  16. 2025π*0.6: a VLA That Learns From Experiencecited by 12 here

The 24 most cited of 311.

Browse

30 of 109, ranked 6–35

  1. 6Evaluating Uncertainty and Quality of Vision-Language-Action-enabled Robots20257
  2. 7CycleVLA: Proactive Self-Correcting Vision-Language-Action Models via Subtask Backtracking and Minimum Bayes Risk Decoding20265
  3. 8When Vision Overrides Language: Evaluating and Mitigating Counterfactual Failures in VLAs20265
  4. 9CaP-X: A Framework for Benchmarking and Improving Coding Agents for Robot Manipulation20264
  5. 10FPC-VLA: A Vision-Language-Action Framework with a Supervisor for Failure Prediction and Correction20254
  6. 11Stable Language Guidance for Vision-Language-Action Models20263
  7. 12LIBERO-X: Robustness Litmus for Vision-Language-Action Models20263
  8. 13Scaling Verification Can Be More Effective than Scaling Policy Learning for Vision-Language-Action Alignment20263
  9. 14Confidence Calibration in Vision-Language-Action Models20253
  10. 15Diagnose, Correct, and Learn from Manipulation Failures via Visual Symbols20253
All 30
  1. 16VLA-Arena: An Open-Source Framework for Benchmarking Vision-Language-Action Models20253
  2. 17See, Plan, Rewind: Progress-Aware Vision-Language-Action Models for Robust Robotic Manipulation20262
  3. 18LIBERO-Para: A Diagnostic Benchmark and Metrics for Paraphrase Robustness in VLA Models20262
  4. 19RoVer: Robot Reward Model as Test-Time Verifier for Vision-Language-Action Model20252
  5. 20NEBULA: Do We Evaluate Vision-Language-Action Agents Correctly?20252
  6. 21UAOR: Uncertainty-aware Observation Reinjection for Vision-Language-Action Models20261
  7. 22Restoring Linguistic Grounding in VLA Models via Train-Free Attention Recalibration20261
  8. 23RoboClaw: An Agentic Framework for Scalable Long-Horizon Robotic Tasks20261
  9. 24Temporal Difference Calibration in Sequential Tasks: Application to Vision-Language-Action Models20261
  10. 25FATE-VLA:Failue-aware test generation for vision-language-action models20261
  11. 26Back to the Familiar Future: Failure Recovery for VLA Policies via Pre-Imagined Milestone Selection20261
  12. 27ReCoVLA: VLM-Guided Reward Compilation for Failure Recovery in Vision-Language-Action Policies20261
  13. 28LIBERO-Occ: Evaluating and Improving Vision-Language-Action Models under Scene-Induced Occlusion via Viewpoint Imagination20261
  14. 29When Does Language Matter? Multilingual Instructions Reveal Step-wise Language Sensitivity in Vision-Language-Action Models20261
  15. 30Uncertainty Quantification for Flow-Based Vision-Language-Action Models20261
  16. 31MINERVA: How Small Can a Manipulation Policy Be and Still Solve LIBERO?20261
  17. 32SwitchVLA: Execution-Aware Task Switching for Vision-Language-Action Models20251
  18. 33Do What? Teaching Vision-Language-Action Models to Reject the Impossible20251
  19. 34RoboChemist: Long-Horizon and Safety-Compliant Robotic Chemical Experimentation20251
  20. 35Bring the Apple, Not the Sofa: Impact of Irrelevant Context in Embodied AI Commands on VLA Models20251