Robots That Know What to Ask: Recovering Misaligned Rewards through Targeted Explanations

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Merker, Helena, Walker, Nick, Bobu, Andreea
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918517586001920
author Merker, Helena
Walker, Nick
Bobu, Andreea
author_facet Merker, Helena
Walker, Nick
Bobu, Andreea
contents Learning reward functions from demonstrations assumes that demonstrations provide adequate supervision over all features -- or task-relevant aspects of behavior. In practice, demonstrations are often imperfect: humans may under-emphasize certain features due to cognitive load or physical difficulty, or the training regime may fail to sufficiently cover all relevant situations. In either case, important features may be underspecified, leading to ambiguity in the learned reward function and misaligned behavior at deployment. We propose a framework that detects such underspecified features and actively solicits targeted corrective demonstrations. Our key insight is that demonstrations implicitly reveal which features are well specified: features that are consistently optimized show little variation across demonstrations, while features that are underspecified vary widely. We leverage this statistical signal to infer which features may have been insufficiently demonstrated. The robot then explains which features it is uncertain about in natural language and queries for demonstrations that explicitly address the identified gaps. We evaluate our approach in a simulated tabletop manipulation domain and in a user study with a real Franka robot. Targeted, explanation-guided queries significantly improve reward recovery compared to random querying and passive data collection, reducing ambiguity that would otherwise persist in learning from imperfect demonstrations.
format Preprint
id arxiv_https___arxiv_org_abs_2605_22986
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Robots That Know What to Ask: Recovering Misaligned Rewards through Targeted Explanations
Merker, Helena
Walker, Nick
Bobu, Andreea
Robotics
Artificial Intelligence
Human-Computer Interaction
Machine Learning
Learning reward functions from demonstrations assumes that demonstrations provide adequate supervision over all features -- or task-relevant aspects of behavior. In practice, demonstrations are often imperfect: humans may under-emphasize certain features due to cognitive load or physical difficulty, or the training regime may fail to sufficiently cover all relevant situations. In either case, important features may be underspecified, leading to ambiguity in the learned reward function and misaligned behavior at deployment. We propose a framework that detects such underspecified features and actively solicits targeted corrective demonstrations. Our key insight is that demonstrations implicitly reveal which features are well specified: features that are consistently optimized show little variation across demonstrations, while features that are underspecified vary widely. We leverage this statistical signal to infer which features may have been insufficiently demonstrated. The robot then explains which features it is uncertain about in natural language and queries for demonstrations that explicitly address the identified gaps. We evaluate our approach in a simulated tabletop manipulation domain and in a user study with a real Franka robot. Targeted, explanation-guided queries significantly improve reward recovery compared to random querying and passive data collection, reducing ambiguity that would otherwise persist in learning from imperfect demonstrations.
title Robots That Know What to Ask: Recovering Misaligned Rewards through Targeted Explanations
topic Robotics
Artificial Intelligence
Human-Computer Interaction
Machine Learning
url https://arxiv.org/abs/2605.22986