Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Tasks

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wu, Zhaofeng, Qiu, Linlu, Ross, Alexis, Akyürek, Ekin, Chen, Boyuan, Wang, Bailin, Kim, Najoung, Andreas, Jacob, Kim, Yoon
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914732554846208
author Wu, Zhaofeng
Qiu, Linlu
Ross, Alexis
Akyürek, Ekin
Chen, Boyuan
Wang, Bailin
Kim, Najoung
Andreas, Jacob
Kim, Yoon
author_facet Wu, Zhaofeng
Qiu, Linlu
Ross, Alexis
Akyürek, Ekin
Chen, Boyuan
Wang, Bailin
Kim, Najoung
Andreas, Jacob
Kim, Yoon
contents The impressive performance of recent language models across a wide range of tasks suggests that they possess a degree of abstract reasoning skills. Are these skills general and transferable, or specialized to specific tasks seen during pretraining? To disentangle these effects, we propose an evaluation framework based on "counterfactual" task variants that deviate from the default assumptions underlying standard tasks. Across a suite of 11 tasks, we observe nontrivial performance on the counterfactual variants, but nevertheless find that performance substantially and consistently degrades compared to the default conditions. This suggests that while current LMs may possess abstract task-solving skills to an extent, they often also rely on narrow, non-transferable procedures for task-solving. These results motivate a more careful interpretation of language model performance that teases apart these aspects of behavior.
format Preprint
id arxiv_https___arxiv_org_abs_2307_02477
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Tasks
Wu, Zhaofeng
Qiu, Linlu
Ross, Alexis
Akyürek, Ekin
Chen, Boyuan
Wang, Bailin
Kim, Najoung
Andreas, Jacob
Kim, Yoon
Computation and Language
Artificial Intelligence
The impressive performance of recent language models across a wide range of tasks suggests that they possess a degree of abstract reasoning skills. Are these skills general and transferable, or specialized to specific tasks seen during pretraining? To disentangle these effects, we propose an evaluation framework based on "counterfactual" task variants that deviate from the default assumptions underlying standard tasks. Across a suite of 11 tasks, we observe nontrivial performance on the counterfactual variants, but nevertheless find that performance substantially and consistently degrades compared to the default conditions. This suggests that while current LMs may possess abstract task-solving skills to an extent, they often also rely on narrow, non-transferable procedures for task-solving. These results motivate a more careful interpretation of language model performance that teases apart these aspects of behavior.
title Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Tasks
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2307.02477