One Word, Different Action: A Real-Robot Benchmark for Language-Conditioned Embodied Reasoning

2026-09-04

AI Summary

研究通过设计一种真实机器人基准测试，评估了语言条件下的具身推理能力，特别是在多约束情况下的决策不变性和敏感性。

Abstract

Natural-language instruction changes can directly alter robot behavior. A reliable embodied system should preserve its action when the task is unchanged and update it correctly when the task itself changes. We introduce One Word, Different Action, a real-robot benchmark built on physical decision states and executable actions, using task-preserving and task-changing instruction pairs to jointly evaluate Decision Invariance and Decision Sensitivity, with further evaluation under multi-constraint reasoning and real-RGB grounding. Experiments show that modern models are near saturation on single-constraint instruction changes, yet several models degrade noticeably when multiple task constraints must be integrated into one executable decision. These results suggest that the more salient remaining challenge is no longer recognizing an isolated instruction change, but reliably composing multiple task requirements into a correct robot action decision.

Problem

• Natural-language instruction
• robot behavior
• Decision Invariance
• Decision Sensitivity
• multi-constraint reasoning

Innovation

• Decision Invariance
• Decision Sensitivity
• Multi-constraint Reasoning