Emergence and Localisation of Semantic Role Circuits in LLMs

TL;DR

This paper investigates how large language models internally represent semantic roles, revealing compact, causally isolated circuits that develop gradually and partially transfer across models and scales.

Contribution

It introduces a novel method combining role-cross minimal pairs, emergence analysis, and cross-model comparison to characterize semantic role circuits in LLMs.

Findings

Highly concentrated semantic role circuits (89-94% attribution within 28 nodes)

Gradual structural refinement of semantic circuits, bypassing localised circuits in larger models

Moderate cross-scale component overlap (24-59%) with high spectral similarity

Abstract

Despite displaying semantic competence, large language models' internal mechanisms that ground abstract semantic structure remain insufficiently characterised. We propose a method integrating role-cross minimal pairs, temporal emergence analysis, and cross-model comparison to study how LLMs implement semantic roles. Our analysis uncovers: (i) highly concentrated circuits (89-94% attribution within 28 nodes); (ii) gradual structural refinement rather than phase transitions, with larger models sometimes bypassing localised circuits; and (iii) moderate cross-scale conservation (24-59% component overlap) alongside high spectral similarity. These findings suggest that LLMs form compact, causally isolated mechanisms for abstract semantic structure, and these mechanisms exhibit partial transfer across scales and architectures.