Large Language Models -- the Future of Fundamental Physics?

TL;DR

This paper explores using Large Language Models, specifically Qwen2.5, for analyzing cosmological data, demonstrating their potential to outperform traditional methods in generating and interpreting large-scale structure maps.

Contribution

It introduces Lightcone LLM (L3M), a novel approach combining LLMs with connector networks for cosmological data analysis, showing improved performance over standard methods.

Findings

L3M outperforms standard initialization in cosmological tasks.

L3M compares favorably with dedicated networks of similar size.

Qwen2.5 enables effective out-of-domain transfer learning for cosmology.

Abstract

For many fundamental physics applications, transformers, as the state of the art in learning complex correlations, benefit from pretraining on quasi-out-of-domain data. The obvious question is whether we can exploit Large Language Models, requiring proper out-of-domain transfer learning. We show how the Qwen2.5 LLM can be used to analyze and generate SKA data, specifically 3D maps of the cosmological large-scale structure for a large part of the observable Universe. We combine the LLM with connector networks and show, for cosmological parameter regression and lightcone generation, that this Lightcone LLM (L3M) with Qwen2.5 weights outperforms standard initialization and compares favorably with dedicated networks of matching size.