Preference-Conditioned Language-Guided Abstraction
HRI ’24, March 11–14, 2024, Boulder, CO, USA
Main Authors: | , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
ACM
2024
|
Online Access: | https://hdl.handle.net/1721.1/154050 |
_version_ | 1811082272924762112 |
---|---|
author | Peng, Andi Bobu, Andreea Li, Belinda Z. Sumers, Theodore R. Sucholutsky, Ilia Kumar, Nishanth Griffiths, Thomas L. Shah, Julie A. |
author_facet | Peng, Andi Bobu, Andreea Li, Belinda Z. Sumers, Theodore R. Sucholutsky, Ilia Kumar, Nishanth Griffiths, Thomas L. Shah, Julie A. |
author_sort | Peng, Andi |
collection | MIT |
description | HRI ’24, March 11–14, 2024, Boulder, CO, USA |
first_indexed | 2024-09-23T12:00:29Z |
format | Article |
id | mit-1721.1/154050 |
institution | Massachusetts Institute of Technology |
language | English |
last_indexed | 2024-09-23T12:00:29Z |
publishDate | 2024 |
publisher | ACM |
record_format | dspace |
spelling | mit-1721.1/1540502024-09-19T05:26:23Z Preference-Conditioned Language-Guided Abstraction Peng, Andi Bobu, Andreea Li, Belinda Z. Sumers, Theodore R. Sucholutsky, Ilia Kumar, Nishanth Griffiths, Thomas L. Shah, Julie A. HRI ’24, March 11–14, 2024, Boulder, CO, USA Learning from demonstrations is a common way for users to teach robots, but it is prone to spurious feature correlations. Recent work constructs state abstractions, i.e. visual representations containing task-relevant features, from language as a way to perform more generalizable learning. However, these abstractions also depend on a user's preference for what matters in a task, which may be hard to describe or infeasible to exhaustively specify using language alone. How do we construct abstractions to capture these latent preferences? We observe that how humans behave reveals how they see the world. Our key insight is that changes in human behavior inform us that there are differences in preferences for how humans see the world, i.e. their state abstractions. In this work, we propose using language models (LMs) to query for those preferences directly given knowledge that a change in behavior has occurred. In our framework, we use the LM in two ways: first, given a text description of the task and knowledge of behavioral change between states, we query the LM for possible hidden preferences; second, given the most likely preference, we query the LM to construct the state abstraction. In this framework, the LM is also able to ask the human directly when uncertain about its own estimate. We demonstrate our framework's ability to construct effective preference-conditioned abstractions in simulated experiments, a user study, as well as on a real Spot robot performing mobile manipulation tasks. 2024-04-03T18:12:14Z 2024-04-03T18:12:14Z 2024-03-11 2024-04-01T07:45:33Z Article http://purl.org/eprint/type/ConferencePaper 979-8-4007-0322-5 https://hdl.handle.net/1721.1/154050 Peng, Andi, Bobu, Andreea, Li, Belinda Z., Sumers, Theodore R., Sucholutsky, Ilia et al. 2024. "Preference-Conditioned Language-Guided Abstraction." PUBLISHER_CC en 10.1145/3610977.3634930 Creative Commons Attribution https://creativecommons.org/licenses/by/4.0/ The author(s) application/pdf ACM ACM |
spellingShingle | Peng, Andi Bobu, Andreea Li, Belinda Z. Sumers, Theodore R. Sucholutsky, Ilia Kumar, Nishanth Griffiths, Thomas L. Shah, Julie A. Preference-Conditioned Language-Guided Abstraction |
title | Preference-Conditioned Language-Guided Abstraction |
title_full | Preference-Conditioned Language-Guided Abstraction |
title_fullStr | Preference-Conditioned Language-Guided Abstraction |
title_full_unstemmed | Preference-Conditioned Language-Guided Abstraction |
title_short | Preference-Conditioned Language-Guided Abstraction |
title_sort | preference conditioned language guided abstraction |
url | https://hdl.handle.net/1721.1/154050 |
work_keys_str_mv | AT pengandi preferenceconditionedlanguageguidedabstraction AT bobuandreea preferenceconditionedlanguageguidedabstraction AT libelindaz preferenceconditionedlanguageguidedabstraction AT sumerstheodorer preferenceconditionedlanguageguidedabstraction AT sucholutskyilia preferenceconditionedlanguageguidedabstraction AT kumarnishanth preferenceconditionedlanguageguidedabstraction AT griffithsthomasl preferenceconditionedlanguageguidedabstraction AT shahjuliea preferenceconditionedlanguageguidedabstraction |