RobotValues: Evaluating Household Robots When Human Values Conflict
Abstract
RobotValues benchmark evaluates household robot planners in value-conflict scenarios, revealing that vision-language models exhibit default value preferences and struggle to override them when instructed to prioritize conflicting values.
While household robots are often evaluated based on task completion, everyday domestic environments involve value-conflicting situations in which robots are expected to choose actions that prioritize other values than task success, such as human autonomy, efficiency, or social appropriateness. Yet, there are no benchmarks for evaluating robots' value preferences in such scenarios. We introduce RobotValues, a benchmark to evaluate household robot planners in 10K value-conflict scenarios. Each instance consists of a realistic household image with multiple plausible robot actions that prioritize different human values. We construct RobotValues through LLM-assisted scenario generation, stakeholder-grounded value extraction, image generation and automatic quality control. Using RobotValues we evaluate VLMs used in robotics and find that models exhibit default value preferences, including safety and accommodation, while underselecting privacy-prioritizing actions. When the models are instructed to prioritize specific values that conflict with their own preferences, they often fail to override their default actions, choosing incorrect actions for 80% of the time. These findings suggest that household robot evaluation should measure not only task completion or safety compliance, but also whether robots can choose among plausible actions when human values conflict.
Community
Seems like a very timely work, nice job!
RobotValues: Evaluating Household Robots When Human Values Conflict
Important angle. Framing value conflicts as a first-class evaluation target feels overdue, and the privacy underselection finding is especially striking. The 80% failure rate at overriding default preferences even when explicitly instructed is a real concern. Great work!
The benchmark's construction through LLM-assisted scenario generation paired with stakeholder-grounded value extraction offers a structured way to surface value conflicts that standard task-success metrics overlook. The reported finding that VLMs fail to override default preferences in 80% of instructed cases highlights a persistent misalignment between model behavior and explicit human-value directives.
How might the automatic quality-control step affect the diversity of value conflicts captured across the 10K scenarios, particularly for less-represented values such as privacy?
I made a podcast on it with ResearchPod, it makes it easy to get the key concepts on the go:
https://researchpod.app/episode/774a4b4d-0095-48d9-84fd-2620c0d83d03
Get this paper in your agent:
hf papers read 2606.03312 Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash Models citing this paper 0
No model linking this paper
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 0
No Space linking this paper
Collections including this paper 0
No Collection including this paper