Inducing Causal World Models in LLMs for Zero-Shot Physical Reasoning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Sharma, Aditya, Gupta, Ananya, Wang, Chengyu, Adebayo, Chiamaka, Kowalski, Jakub
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909976141758464
author Sharma, Aditya
Gupta, Ananya
Wang, Chengyu
Adebayo, Chiamaka
Kowalski, Jakub
author_facet Sharma, Aditya
Gupta, Ananya
Wang, Chengyu
Adebayo, Chiamaka
Kowalski, Jakub
contents Large Language Models (LLMs), despite their advanced linguistic capabilities, fundamentally lack an intuitive understanding of physical dynamics, which limits their effectiveness in real-world scenarios that require causal reasoning. In this paper, we introduce Causal World Model Induction (CWMI), a novel framework designed to embed an explicit model of causal physics within an LLM. Our approach incorporates a dedicated Causal Physics Module (CPM) and a new training objective called Causal Intervention Loss, encouraging the model to learn cause-and-effect relationships from multimodal data. By training the model to predict the outcomes of hypothetical interventions instead of merely capturing statistical correlations, CWMI develops a robust internal representation of physical laws. Experimental results show that CWMI significantly outperforms state-of-the-art LLMs on zero-shot physical reasoning tasks, including the PIQA benchmark and our newly proposed PhysiCa-Bench dataset. These findings demonstrate that inducing a causal world model is a critical step toward more reliable and generalizable AI systems.
format Preprint
id arxiv_https___arxiv_org_abs_2507_19855
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Inducing Causal World Models in LLMs for Zero-Shot Physical Reasoning
Sharma, Aditya
Gupta, Ananya
Wang, Chengyu
Adebayo, Chiamaka
Kowalski, Jakub
Machine Learning
Human-Computer Interaction
68T05, 68T07, 68T40
I.2.6; I.2.9; I.2.7; I.2.10; H.5.2
Large Language Models (LLMs), despite their advanced linguistic capabilities, fundamentally lack an intuitive understanding of physical dynamics, which limits their effectiveness in real-world scenarios that require causal reasoning. In this paper, we introduce Causal World Model Induction (CWMI), a novel framework designed to embed an explicit model of causal physics within an LLM. Our approach incorporates a dedicated Causal Physics Module (CPM) and a new training objective called Causal Intervention Loss, encouraging the model to learn cause-and-effect relationships from multimodal data. By training the model to predict the outcomes of hypothetical interventions instead of merely capturing statistical correlations, CWMI develops a robust internal representation of physical laws. Experimental results show that CWMI significantly outperforms state-of-the-art LLMs on zero-shot physical reasoning tasks, including the PIQA benchmark and our newly proposed PhysiCa-Bench dataset. These findings demonstrate that inducing a causal world model is a critical step toward more reliable and generalizable AI systems.
title Inducing Causal World Models in LLMs for Zero-Shot Physical Reasoning
topic Machine Learning
Human-Computer Interaction
68T05, 68T07, 68T40
I.2.6; I.2.9; I.2.7; I.2.10; H.5.2
url https://arxiv.org/abs/2507.19855