Selective Perception: Learning Concise State Descriptions for Language Model Actors

Kolby Nottingham, Yasaman Razeghi, Kyungmin Kim, Jb Lanier, Pierre Baldi, Roy Fox, Sameer Singh


Abstract
The latest large language models (LMs) support increasingly longer contexts. While this trend permits using substantial amounts of text with SOTA LMs, requiring these large LMs to process potentially redundant or irrelevant data needlessly increases inference time and cost. To remedy this problem, we propose BLINDER, a method that leverages a small finetuned LM to sample the minimal set of input features that maximizes the performance of a downstream LM. BLINDER trains an LM with a value head to estimate the likelihood of optimal outputs from a downstream LM given an input. We evaluate BLINDER on embodied decision making tasks with notoriously verbose state descriptions: NetHack and robot planning. BLINDER reduces the length of LM actor input by 87% and 99% while improving task success rates by 158% and 54% on NetHack and robot planning respectively which represents substantial inference cost savings while actually increasing performance.
Anthology ID:
2024.naacl-short.29
Volume:
Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 2: Short Papers)
Month:
June
Year:
2024
Address:
Mexico City, Mexico
Editors:
Kevin Duh, Helena Gomez, Steven Bethard
Venue:
NAACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
327–341
Language:
URL:
https://aclanthology.org/2024.naacl-short.29
DOI:
10.18653/v1/2024.naacl-short.29
Bibkey:
Cite (ACL):
Kolby Nottingham, Yasaman Razeghi, Kyungmin Kim, Jb Lanier, Pierre Baldi, Roy Fox, and Sameer Singh. 2024. Selective Perception: Learning Concise State Descriptions for Language Model Actors. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 2: Short Papers), pages 327–341, Mexico City, Mexico. Association for Computational Linguistics.
Cite (Informal):
Selective Perception: Learning Concise State Descriptions for Language Model Actors (Nottingham et al., NAACL 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.naacl-short.29.pdf