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

Zhaofeng Wu, Linlu Qiu, Alexis Ross, Ekin Akyürek, Boyuan Chen +4 more
7/5/2023
cs.CLcs.AI

No official code found. Best implementation: ZhaofengWu/counterfactual-evaluation (61 stars, MIT, last commit 2025-05-19).

Abstract

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.

View on arXivView PDF

Code Implementations(1)

ZhaofengWu/counterfactual-evaluationCommunity100%
MIT
616Shell, PythonJun 29, 20231 years agoMIT

Similar Papers

Evidence from counterfactual tasks supports emergent analogical reasoning in large language models

87% match1 repo22024

Measuring Faithfulness in Chain-of-Thought Reasoning

86% match2023

Generative Models as a Complex Systems Science: How can we make sense of large language model behavior?

86% match2023

What do Language Models Learn and When? The Implicit Curriculum Hypothesis

86% match2026

Large GPT-like Models are Bad Babies: A Closer Look at the Relationship between Linguistic Competence and Psycholinguistic Measures

86% match2023

CLOMO: Counterfactual Logical Modification with Large Language Models

85% match1 repo42023

Cite this paper

@article{wu2023reasoning,
  title  = {Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Tasks},
  author = {Zhaofeng Wu and Linlu Qiu and Alexis Ross and Ekin Akyürek and Boyuan Chen and Bailin Wang and Najoung Kim and Jacob Andreas and Yoon Kim},
  year   = {2023},
  eprint = {2307.02477},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2307.02477v3}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2307.02477.svg)](https://codeofpaper.com/paper/2307.02477)

Discussion