A Peek into Token Bias: Large Language Models Are Not Yet Genuine Reasoners

TL;DR

This paper investigates whether large language models genuinely reason or rely on token bias, revealing that most models depend on superficial patterns rather than true logical reasoning abilities.

Contribution

The study introduces a hypothesis-testing framework with synthetic datasets to systematically evaluate token bias versus genuine reasoning in LLMs.

Findings

Most LLMs struggle with logical reasoning tasks.

Performance depends on superficial token patterns, not reasoning.

Concerns raised about models' generalization abilities.

Abstract

This study introduces a hypothesis-testing framework to assess whether large language models (LLMs) possess genuine reasoning abilities or primarily depend on token bias. We go beyond evaluating LLMs on accuracy; rather, we aim to investigate their token bias in solving logical reasoning tasks. Specifically, we develop carefully controlled synthetic datasets, featuring conjunction fallacy and syllogistic problems. Our framework outlines a list of hypotheses where token biases are readily identifiable, with all null hypotheses assuming genuine reasoning capabilities of LLMs. The findings in this study suggest, with statistical guarantee, that most LLMs still struggle with logical reasoning. While they may perform well on classic problems, their success largely depends on recognizing superficial patterns with strong token bias, thereby raising concerns about their actual reasoning and generalization abilities. Codes and data are open-sourced at https://github.com/bowen-upenn/llm_token_bias.