# When Should a Failing Robot Ask? Initiating Corrective Human-Robot Dialogue from Audited Sensor Evidence

arXiv: https://arxiv.org/abs/2609.21942 · HTML: https://mlsanity.com/paper/2609.21942/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 25 of its 38 references that are in the map.
- First posted on arXiv 18 Sep 2026.

## Builds on (most cited first, 25 of 25)

- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [On Calibration of Modern Neural Networks](https://mlsanity.com/paper/1706.04599/) (2017)
- [LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learning](https://mlsanity.com/paper/2306.03310/) (2023)
- [Large Language Models Are Not Robust Multiple Choice Selectors](https://mlsanity.com/paper/2309.03882/) (2023)
- [Robots That Ask For Help: Uncertainty Alignment for Large Language Model Planners](https://mlsanity.com/paper/2307.01928/) (2023)
- [REFLECT: Summarizing Robot Experiences for Failure Explanation and Correction](https://mlsanity.com/paper/2306.15724/) (2023)
- [Cosmos 3: Omnimodal World Models for Physical AI](https://mlsanity.com/paper/2606.02800/) (2026)
- [AHA: A Vision-Language-Model for Detecting and Reasoning Over Failures in Robotic Manipulation](https://mlsanity.com/paper/2410.00371/) (2024)
- [Yell At Your Robot: Improving On-the-Fly from Language Corrections](https://mlsanity.com/paper/2403.12910/) (2024)
- [Correcting Robot Plans with Natural Language Feedback](https://mlsanity.com/paper/2204.05186/) (2022)
- [Code-as-Monitor: Constraint-aware Visual Programming for Reactive and Proactive Robotic Failure Detection](https://mlsanity.com/paper/2412.04455/) (2024)
- [PersonaPlex: Voice and Role Control for Full Duplex Conversational Speech Models](https://mlsanity.com/paper/2602.06053/) (2026)
- [RACER: Rich Language-Guided Failure Recovery Policies for Imitation Learning](https://mlsanity.com/paper/2409.14674/) (2024)
- [Distilling and Retrieving Generalizable Knowledge for Robot Manipulation via Language Corrections](https://mlsanity.com/paper/2311.10678/) (2023)
- [HiL-Bench (Human-in-Loop Benchmark): Do Agents Know When to Ask for Help?](https://mlsanity.com/paper/2604.09408/) (2026)
- [Diagnose, Correct, and Learn from Manipulation Failures via Visual Symbols](https://mlsanity.com/paper/2512.02787/) (2025)
- [When Robots Should Say "I Don't Know": Benchmarking Abstention in Embodied Question Answering](https://mlsanity.com/paper/2512.04597/) (2025)
- [Learning Actionable Manipulation Recovery via Counterfactual Failure Synthesis](https://mlsanity.com/paper/2603.13528/) (2026)
- [A Survey on Dialogue Management in Human-Robot Interaction](https://mlsanity.com/paper/2307.10897/) (2023)
- [InteLiPlan: An Interactive Lightweight LLM-Based Planner for Domestic Robot Autonomy](https://mlsanity.com/paper/2409.14506/) (2024)
- [SCOUT: A Situated and Multi-Modal Human-Robot Dialogue Corpus](https://mlsanity.com/paper/2411.12844/) (2024)
- [The Yes-Man Syndrome: Benchmarking Abstention in Embodied Robotic Agents](https://mlsanity.com/paper/2605.20544/) (2026)
- [Ask When It Pays: Cost-Aware Open-Ended Interaction for Instance Goal Navigation](https://mlsanity.com/paper/2606.03175/) (2026)
- [ESearch-R1: Learning Cost-Aware MLLM Agents for Interactive Embodied Search via Reinforcement Learning](https://mlsanity.com/paper/2512.18571/) (2025)
- [Benchmarking Interaction, Beyond Policy: a Reproducible Benchmark for Collaborative Instance Object Navigation](https://mlsanity.com/paper/2604.00265/) (2026)

## Abstract (arXiv)

A robot that fails at a task faces the first decision in corrective dialogue: act on its own diagnosis, consult another onboard sensor, or interrupt a person. Choosing well requires knowing how much the robot's sensors reveal about the cause and how reliable the robot's own diagnosis is. We build a simulated benchmark in which every failure's true cause is known, because we injected it, and measure what each sensor reveals, with explicit checks against data leakage. Some failures are diagnosable from camera images; others only from the robot's force data (0.99 from force data, no image method above 0.55). We then test six open vision-language models. Their behavior tracks the surface of the prompt, not the evidence: moving the refusal option from last to first in the answer list collapses refusal rates from 78-100% to 0-6% in three of the six swept model-and-family pairs. Accuracy from frames stays at or below a majority-class baseline under every prompt variant, with or without worked examples, and stated confidence carries no information about correctness. Handing the same models the force data as ten lines of text produces the first above-baseline diagnoses, in four of the six models: much of the failure reflects missing sensor data, not missing ability. We pose the choice as a three-action decision problem, act, consult your own sensors, or ask a human, whose optimal policy follows from measured accuracy. The models do not follow it, and their ask rates ignore a fourfold change in question cost. One question to a human still lifts them from that baseline to roughly the answerer's own reliability (0.70-0.81 when they ask). The decision to ask should be tied to measured accuracy and stated costs, not to the model's confidence.