Overcoming Referential Ambiguity in Language-Guided Goal-Conditioned Reinforcement Learning

Abstract

Teaching an agent to perform new tasks using natural language can easily be hindered by ambiguities in interpretation. When a teacher provides an instruction to a learner about an object by referring to its features, the learner can misunderstand the teacher's intentions, for instance if the instruction ambiguously refer to features of the object, a phenomenon called referential ambiguity. We study how two concepts derived from cognitive sciences can help resolve those referential ambiguities: pedagogy (selecting the right instructions) and pragmatism (learning the preferences of the other agents using inductive reasoning). We apply those ideas to a teacher/learner setup with two artificial agents on a simulated robotic task (block-stacking). We show that these concepts improve sample efficiency for training the learner.

Cite

Text

Caselles-Dupré et al. "Overcoming Referential Ambiguity in Language-Guided Goal-Conditioned Reinforcement Learning." NeurIPS 2022 Workshops: LaReL, 2022.

Markdown

[Caselles-Dupré et al. "Overcoming Referential Ambiguity in Language-Guided Goal-Conditioned Reinforcement Learning." NeurIPS 2022 Workshops: LaReL, 2022.](https://mlanthology.org/neuripsw/2022/casellesdupre2022neuripsw-overcoming/)

BibTeX

@inproceedings{casellesdupre2022neuripsw-overcoming,
  title     = {{Overcoming Referential Ambiguity in Language-Guided Goal-Conditioned Reinforcement Learning}},
  author    = {Caselles-Dupré, Hugo and Sigaud, Olivier and Chetouani, Mohamed},
  booktitle = {NeurIPS 2022 Workshops: LaReL},
  year      = {2022},
  url       = {https://mlanthology.org/neuripsw/2022/casellesdupre2022neuripsw-overcoming/}
}