Language models can learn implicit multi-hop reasoning, but only if they have lots of training data
Year: 2025
Type: preprint
Abstract: Implicit reasoning is the ability of a language model to solve multi-hop reasoning tasks in a single forward pass, without chain of thought. We investigate this capability using GPT2-style language models trained from scratch on controlled $k$-hop reasoning datasets ($k = 2, 3, 4$). We show that while such models can indeed learn implicit $k$-hop reasoning, the required training data grows exponentially in $k$, and the required number of transformer layers grows linearly in $k$. We offer a theoretical explanation for why this depth growth is necessary. We further find that the data requirement can be mitigated, but not eliminated, through curriculum learning.
Source: arXiv (Cornell University)
Authors: Yao, Yuekun, Du, Yupei, Zhu, Dawei, Hahn, Michael, Koller, Alexander
Institutions: Saarland University, Utrecht University
Language: English
Topic: Topic Modeling
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W6979233995
Cites:
Cited by:
Related to: