Research map / Vision-language and multimodal / Robot action and navigation
Diagnose, Correct, and Learn from Manipulation Failures via Visual Symbols
Xianchao Zeng, Xinyu Zhou, Youcheng Li, Jiayou Shi, Tianle Li, Liangming Chen, Lei Ren, Yong-Lu 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 |
|---|---|---|---|
| fine-tune on ViFailback dataset (LoRA, 1 epoch) (ablation) | Qwen3-VL-8B · ViFailback-Bench Lite | accuracy (%): 38.33 -> 93.70 | Improved |
| fine-tune on ViFailback dataset (ablation) | Qwen3-VL-8B · ViFailback-Bench Hard | score (%): 33.04 -> 72.64 | Improved |
| increase training trajectories from 1,200 to 4,702 (ablation) | Qwen3-VL-8B · ViFailback-Bench | accuracy (%) | Improved |
| increase training data on visual symbol code generation (ablation) | Qwen3-VL-8B · Generating Visual Symbols' Codes | accuracy (%): up to 38.73 | Improved |
| scale model size within Qwen2.5-VL (3B to 72B) | Qwen2.5-VL · ViFailback-Bench Lite | accuracy (%): 38.10 -> 50.61 | Improved |
| scale Qwen3-VL from 8B to 32B | Qwen3-VL · ViFailback-Bench Lite | accuracy (%): 38.33 -> 47.79 | Improved |
| add ViFailback-8B correction with visual symbols following data (VSF) | pi0.5 (base & symbol) · real-world manipulation | success rate (%): 52.4% -> 73.0% | Improved |
| add ViFailback-8B correction with point-based motion control (PMC) | pi0.5 (base) · real-world manipulation | success rate (%): 50.8% -> 74.6% | Improved |
| add visual symbol-following data to pi0.5 fine-tuning without correction (ablation) | pi0.5 · real-world manipulation | success rate (%): 50.8% -> 52.4% | No clear effect |
| PMC instead of VSF for executing guidance (ablation) | ViFailback-8B system · real-world manipulation | success rate (%): 73.0% -> 74.6% | 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.
- RoboBrain2.0-32B — RoboBrain: A Unified Brain Model for Robotic Manipulation from Abstract to Concrete (CVPR 2025) · better on 2 of 2 shared results
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
- AHA: A Vision-Language-Model for Detecting and Reasoning Over Failures in Robotic Manipulation (ICLR 2025 · similarity 0.92)
- SAFE: Multitask Failure Detection for Vision-Language-Action Models (NeurIPS 2025 · similarity 0.91)
- Failing Forward: Adaptive Failure-Informed Learning for Vision-Language-Action Models (NeurIPS 2026 · similarity 0.90)
- RoboProcessBench: Benchmarking Process-Aware Understanding in Visual-Language Robotic Manipulation (NeurIPS 2026 · similarity 0.90)
- CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Models (CVPR 2025 · similarity 0.89)
- VLA-REPLICA: A Low-Cost, Reproducible Benchmark for Real-World Evaluation of Vision-Language-Action Models (NeurIPS 2026 · similarity 0.89)
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.
