Source record

Published by arXiv cs.RO (T1 source). The original is at https://arxiv.org/abs/2608.04309.

Pipeline notes

The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.

A structured LLM architecture for zero-shot human-robot coordination in a cooperative construction task decomposes decision-making into theory-of-mind inference, hierarchical planning, conversation interpretation, action verification, and feedback-based replanning. Human-participant experiments showed fewer interaction steps and higher trust than ablated and reinforcement-learning baselines.

Concrete five-part decomposition of LLM reasoning for team coordination, with empirical human-participant data showing efficiency and trust gains over baselines.

Cited by

No citations on record.