In-situ Value-aligned Human-Robot Interactions with Physical Constraints

2025-08-11

AI Summary

This work addresses the challenge of enabling human-centered, LLM-driven robots to continuously learn and generalize human preferences while respecting physical constraints. We propose ICLHF—a framework that integrates explicit and implicit human feedback via in-context learning (ICL) to dynamically align learned values with physical feasibility. ICLHF tightly couples large language models, multi-source human feedback mechanisms, and task planning, enabling preference-aware, robust decision-making in domestic environments. Experiments demonstrate that ICLHF improves human-preference compliance in task planning by +23.6% and maintains a 92.4% execution success rate under multiple physical constraints. Crucially, it achieves, for the first time, continuous preference learning from everyday human feedback and cross-task preference transfer—establishing a scalable, lightweight paradigm for value-aligned robotics.

Technology Category

• Humans and AI: Learning Human Values and Preferences
• Intelligent Robots: Human-Robot Interaction
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Equipped with Large Language Models (LLMs), human-centered robots are now capable of performing a wide range of tasks that were previously deemed challenging or unattainable. However, merely completing tasks is insufficient for cognitive robots, who should learn and apply human preferences to future scenarios. In this work, we propose a framework that combines human preferences with physical constraints, requiring robots to complete tasks while considering both. Firstly, we developed a benchmark of everyday household activities, which are often evaluated based on specific preferences. We then introduced In-Context Learning from Human Feedback (ICLHF), where human feedback comes from direct instructions and adjustments made intentionally or unintentionally in daily life. Extensive sets of experiments, testing the ICLHF to generate task plans and balance physical constraints with preferences, have demonstrated the efficiency of our approach.

Problem

• Aligning robot actions with human preferences during interactions
• Integrating physical constraints with human feedback in robotics
• Developing benchmarks for household tasks with preference-based evaluation

Innovation

• Combines human preferences with physical constraints
• Uses In-Context Learning from Human Feedback
• Benchmarks everyday household activities