TY - GEN
T1 - Behavior Cloned Transformers are Neurosymbolic Reasoners
AU - Wang, Ruoyao
AU - Jansen, Peter
AU - Côté, Marc Alexandre
AU - Ammanabrolu, Prithviraj
N1 - Funding Information:
This work supported in part by National Science Foundation (NSF) award #1815948 to PJ, and the Allen Institute for Artificial Intelligence (AI2).
Publisher Copyright:
© 2023 Association for Computational Linguistics.
PY - 2023
Y1 - 2023
N2 - In this work, we explore techniques for augmenting interactive AI agents with information from symbolic modules, much like humans use tools like calculators and GPS systems to assist with arithmetic and navigation. We test our agent's abilities in text games-challenging benchmarks for evaluating the multi-step reasoning abilities of game agents in grounded, language-based environments. Our experimental study indicates that injecting the actions from these symbolic modules into the action space of a behavior cloned transformer agent increases performance on four text game benchmarks that test arithmetic, navigation, sorting, and common sense reasoning by an average of 22%, allowing an agent to reach the highest possible performance on unseen games. This action injection technique is easily extended to new agents, environments, and symbolic modules.
AB - In this work, we explore techniques for augmenting interactive AI agents with information from symbolic modules, much like humans use tools like calculators and GPS systems to assist with arithmetic and navigation. We test our agent's abilities in text games-challenging benchmarks for evaluating the multi-step reasoning abilities of game agents in grounded, language-based environments. Our experimental study indicates that injecting the actions from these symbolic modules into the action space of a behavior cloned transformer agent increases performance on four text game benchmarks that test arithmetic, navigation, sorting, and common sense reasoning by an average of 22%, allowing an agent to reach the highest possible performance on unseen games. This action injection technique is easily extended to new agents, environments, and symbolic modules.
UR - http://www.scopus.com/inward/record.url?scp=85159861401&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85159861401&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:85159861401
T3 - EACL 2023 - 17th Conference of the European Chapter of the Association for Computational Linguistics, Proceedings of the Conference
SP - 2769
EP - 2780
BT - EACL 2023 - 17th Conference of the European Chapter of the Association for Computational Linguistics, Proceedings of the Conference
PB - Association for Computational Linguistics (ACL)
T2 - 17th Conference of the European Chapter of the Association for Computational Linguistics, EACL 2023
Y2 - 2 May 2023 through 4 May 2023
ER -