Research map / Vision-language and multimodal / Robot action and navigation
FLARE: A Failure-Aware Framework for Autonomous Correction and Recovery in Visual-Language Robotic Manipulation
Ganlong Zhao, Zijia Tang, Xingping Chen, Zhanghui Kuang, Ye Tian, Guanbin Li
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add FLARE retry/reset framework on top of π0.5 | π0.5 · RoboMimic 9 tasks | mean success rate: 72.2% -> 84.0% | Improved |
| use FLARE instead of Phoenix | Phoenix · RoboMimic 9 tasks | mean success rate: 57.8% -> 84.0% | Improved |
| add FLARE to π0.5 on Stack D0 | π0.5 · RoboMimic Stack D0 | success rate: 100% -> 100% | No clear effect |
| add FLARE to π0.5 on real robot | π0.5 · Real-world Stack Three Blocks | success rate: 62.5% -> 75.0% | Improved |
| add FLARE to π0.5 on real robot | π0.5 · Real-world Insert U-shaped Block | success rate: 45.0% -> 55.0% | Improved |
| add reset skills to perturbation & bridging (ablation) | FLARE · Coffee and ThreePieceAssembly | success rate: Coffee D1 74% -> 78%; TPA D1 54% -> 58% | Improved |
| add perturbation & bridging retry data to reset-only system (ablation) | FLARE · Coffee D1 | success rate: 64% -> 78% | Improved |
| replace MLLM monitor with human instructions (oracle) (ablation) | FLARE · Coffee and ThreePieceAssembly | success rate: Coffee D1 78% -> 90% | Improved |
| use moderate perturbation magnitude (r=30 deg, t=0.7) (ablation) | FLARE retry augmentation · Coffee D1 | success rate | Improved |
| use excessively large perturbations (ablation) | FLARE retry augmentation · Coffee D1 | success rate | Hurt |
| reset an object needing pose adjustment (pod) vs simple object (lid) | FLARE reset skills · Coffee reset skills | reset success rate: lid 84% vs pod 24% | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Phoenix — Phoenix: A Motion-based Self-Reflection Framework for Fine-grained Robotic Action Correction (CVPR 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Failing Forward: Adaptive Failure-Informed Learning for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.92)
- AHA: A Vision-Language-Model for Detecting and Reasoning Over Failures in Robotic Manipulation (ICLR 2025 · similarity 0.91)
- RePO-VLA: Recovery-Driven Policy Optimization for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.89)
- RoboProcessBench: Benchmarking Process-Aware Understanding in Visual-Language Robotic Manipulation (NeurIPS 2026 · similarity 0.89)
- SAFE: Multitask Failure Detection for Vision-Language-Action Models (NeurIPS 2025 · similarity 0.88)
- Self-Correcting Robot Manipulation via Gaussian-Splatted Foresight (AAAI 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
