# FailBench: How Reliable are VLMs at Judging Robot Task Success?

arXiv: https://arxiv.org/abs/2609.03611 · HTML: https://mlsanity.com/paper/2609.03611/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Embodied and 3D vision-language models, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 47 of its 49 references that are in the map.
- First posted on arXiv 3 Sep 2026.

## Placed in

- Multimodal LLMs / vision-language models / [Embodied and 3D vision-language models](https://mlsanity.com/mllm/embodied-and-3d-vision-language-models/)

## Builds on (most cited first, 30 of 47)

- [GPT-4o System Card](https://mlsanity.com/paper/2410.21276/) (2024)
- [Qwen3-VL Technical Report](https://mlsanity.com/paper/2511.21631/) (2025)
- [Evaluating Real-World Robot Manipulation Policies in Simulation](https://mlsanity.com/paper/2405.05941/) (2024)
- [LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methods](https://mlsanity.com/paper/2412.05579/) (2024)
- [Gemma 4 Technical Report](https://mlsanity.com/paper/2607.02770/) (2026)
- [RH20T: A Comprehensive Robotic Dataset for Learning Diverse Skills in One-Shot](https://mlsanity.com/paper/2307.00595/) (2023)
- [VLA-RL: Towards Masterful and General Robotic Manipulation with Scalable Reinforcement Learning](https://mlsanity.com/paper/2505.18719/) (2025)
- [REFLECT: Summarizing Robot Experiences for Failure Explanation and Correction](https://mlsanity.com/paper/2306.15724/) (2023)
- [AHA: A Vision-Language-Model for Detecting and Reasoning Over Failures in Robotic Manipulation](https://mlsanity.com/paper/2410.00371/) (2024)
- [GRAPE: Generalizing Robot Policy via Preference Alignment](https://mlsanity.com/paper/2411.19309/) (2024)
- [SAFE: Multitask Failure Detection for Vision-Language-Action Models](https://mlsanity.com/paper/2506.09937/) (2025)
- [Unpacking Failure Modes of Generative Policies: Runtime Monitoring of Consistency and Progress](https://mlsanity.com/paper/2410.04640/) (2024)
- [WorldGym: World Model as An Environment for Policy Evaluation](https://mlsanity.com/paper/2506.00613/) (2025)
- [RoboArena: Distributed Real-World Evaluation of Generalist Robot Policies](https://mlsanity.com/paper/2506.18123/) (2025)
- [Code-as-Monitor: Constraint-aware Visual Programming for Reactive and Proactive Robotic Failure Detection](https://mlsanity.com/paper/2412.04455/) (2024)
- [Robometer: Scaling General-Purpose Robotic Reward Models via Trajectory Comparisons](https://mlsanity.com/paper/2603.02115/) (2026)
- [RoboReward: General-Purpose Vision-Language Reward Models for Robotics](https://mlsanity.com/paper/2601.00675/) (2026)
- [Can We Detect Failures Without Failure Data? Uncertainty-Aware Runtime Failure Detection for Imitation Learning Policies](https://mlsanity.com/paper/2503.08558/) (2025)
- [AutoEval: Autonomous Evaluation of Generalist Robot Manipulation Policies in the Real World](https://mlsanity.com/paper/2503.24278/) (2025)
- [Failure Prediction at Runtime for Generative Robot Policies](https://mlsanity.com/paper/2510.09459/) (2025)
- [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)
- [Autonomous Improvement of Instruction Following Skills via Foundation Models](https://mlsanity.com/paper/2407.20635/) (2024)
- [RoboEval: Where Robotic Manipulation Meets Structured and Scalable Evaluation](https://mlsanity.com/paper/2507.00435/) (2025)
- [Robobench: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models as Embodied Brain](https://mlsanity.com/paper/2510.17801/) (2025)
- [Are We on the Right Way to Assessing LLM-as-a-Judge?](https://mlsanity.com/paper/2512.16041/) (2025)
- [Automating Robot Failure Recovery Using Vision-Language Models With Optimized Prompts](https://mlsanity.com/paper/2409.03966/) (2024)
- [Reliable and Scalable Robot Policy Evaluation with Imperfect Simulators](https://mlsanity.com/paper/2510.04354/) (2025)
- [Diagnose, Correct, and Learn from Manipulation Failures via Visual Symbols](https://mlsanity.com/paper/2512.02787/) (2025)
- [I-FailSense: Towards General Robotic Failure Detection with Vision-Language Models](https://mlsanity.com/paper/2509.16072/) (2025)

## Abstract (arXiv)

Vision-Language Models (VLMs) are increasingly used to evaluate robot manipulation outcomes, but existing benchmarks offer limited evidence of cross-domain generalization. We introduce FailBench, a benchmark for robot failure detection comprising 2,197 manipulation attempts across 14 public sources (12 real-world, 2 simulated). In FailBench, 75% of failures occur naturally, and six real-world sources come from non-failure-detection datasets. Evaluating 13 VLM-based detectors, we find the best model achieves only 0.77 mean balanced accuracy. Notably, models fine-tuned for failure detection consistently underperform general-purpose VLMs and their own pretrained baselines. Performance depends heavily on required visual evidence: models approach saturation when outcomes depend on observable object motion, but degrade to near-chance (<0.60 balanced accuracy) on contact-intensive assembly tasks. Error analysis reveals a systematic bias toward predicting success under ambiguous evidence, which persists even with increased reasoning effort. Finally, we show that input-level intervention--spatially localizing and cropping outcome-relevant regions--improves the top detector by 2.4 percentage points without extra training.