UNcommonsense Reasoning: Abductive Reasoning about Uncommon Situations

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Zhao, Wenting, Chiu, Justin T, Hwang, Jena D., Brahman, Faeze, Hessel, Jack, Choudhury, Sanjiban, Choi, Yejin, Li, Xiang Lorraine, Suhr, Alane
Formato: Preprint
Publicado: 2023
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866914778175242240
author Zhao, Wenting
Chiu, Justin T
Hwang, Jena D.
Brahman, Faeze
Hessel, Jack
Choudhury, Sanjiban
Choi, Yejin
Li, Xiang Lorraine
Suhr, Alane
author_facet Zhao, Wenting
Chiu, Justin T
Hwang, Jena D.
Brahman, Faeze
Hessel, Jack
Choudhury, Sanjiban
Choi, Yejin
Li, Xiang Lorraine
Suhr, Alane
contents Language technologies that accurately model the dynamics of events must perform commonsense reasoning. Existing work evaluating commonsense reasoning focuses on making inferences about common, everyday situations. To instead investigate the ability to model unusual, unexpected, and unlikely situations, we explore the task of uncommonsense abductive reasoning. Given a piece of context with an unexpected outcome, this task requires reasoning abductively to generate an explanation that makes the unexpected outcome more likely in the context. To this end, we curate and release a new English language corpus called UNcommonsense. We characterize the performance differences between human explainers and the best-performing large language models, finding that model-enhanced human-written explanations achieve the highest quality by trading off between specificity and diversity. Finally, we experiment with several imitation learning algorithms to train open and accessible language models on this task. When compared with the vanilla supervised fine-tuning approach, these methods consistently reduce lose rates on both common and uncommonsense abductive reasoning judged by human evaluators.
format Preprint
id arxiv_https___arxiv_org_abs_2311_08469
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle UNcommonsense Reasoning: Abductive Reasoning about Uncommon Situations
Zhao, Wenting
Chiu, Justin T
Hwang, Jena D.
Brahman, Faeze
Hessel, Jack
Choudhury, Sanjiban
Choi, Yejin
Li, Xiang Lorraine
Suhr, Alane
Computation and Language
Language technologies that accurately model the dynamics of events must perform commonsense reasoning. Existing work evaluating commonsense reasoning focuses on making inferences about common, everyday situations. To instead investigate the ability to model unusual, unexpected, and unlikely situations, we explore the task of uncommonsense abductive reasoning. Given a piece of context with an unexpected outcome, this task requires reasoning abductively to generate an explanation that makes the unexpected outcome more likely in the context. To this end, we curate and release a new English language corpus called UNcommonsense. We characterize the performance differences between human explainers and the best-performing large language models, finding that model-enhanced human-written explanations achieve the highest quality by trading off between specificity and diversity. Finally, we experiment with several imitation learning algorithms to train open and accessible language models on this task. When compared with the vanilla supervised fine-tuning approach, these methods consistently reduce lose rates on both common and uncommonsense abductive reasoning judged by human evaluators.
title UNcommonsense Reasoning: Abductive Reasoning about Uncommon Situations
topic Computation and Language
url https://arxiv.org/abs/2311.08469