Saved in:
Bibliographic Details
Main Authors: Li, Shenxiong, Rui, Huaxia
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2412.11009
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915063909056512
author Li, Shenxiong
Rui, Huaxia
author_facet Li, Shenxiong
Rui, Huaxia
contents We conducted three experiments to investigate how large language models (LLMs) evaluate posterior probabilities. Our results reveal the coexistence of two modes in posterior judgment among state-of-the-art models: a normative mode, which adheres to Bayes' rule, and a representative-based mode, which relies on similarity -- paralleling human System 1 and System 2 thinking. Additionally, we observed that LLMs struggle to recall base rate information from their memory, and developing prompt engineering strategies to mitigate representative-based judgment may be challenging. We further conjecture that the dual modes of judgment may be a result of the contrastive loss function employed in reinforcement learning from human feedback. Our findings underscore the potential direction for reducing cognitive biases in LLMs and the necessity for cautious deployment of LLMs in critical areas.
format Preprint
id arxiv_https___arxiv_org_abs_2412_11009
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Dual Traits in Probabilistic Reasoning of Large Language Models
Li, Shenxiong
Rui, Huaxia
Artificial Intelligence
Computation and Language
Computers and Society
We conducted three experiments to investigate how large language models (LLMs) evaluate posterior probabilities. Our results reveal the coexistence of two modes in posterior judgment among state-of-the-art models: a normative mode, which adheres to Bayes' rule, and a representative-based mode, which relies on similarity -- paralleling human System 1 and System 2 thinking. Additionally, we observed that LLMs struggle to recall base rate information from their memory, and developing prompt engineering strategies to mitigate representative-based judgment may be challenging. We further conjecture that the dual modes of judgment may be a result of the contrastive loss function employed in reinforcement learning from human feedback. Our findings underscore the potential direction for reducing cognitive biases in LLMs and the necessity for cautious deployment of LLMs in critical areas.
title Dual Traits in Probabilistic Reasoning of Large Language Models
topic Artificial Intelligence
Computation and Language
Computers and Society
url https://arxiv.org/abs/2412.11009