TESTING FOR COHERENCE SUPPRESSION IN FRONTIER LLMs: AN EXPERIMENTAL DESIGN PROPOSAL

Read full text
Authors: Edoardo Sorrentino
Cited by: 2
References: 1

Abstract

This paper proposes a falsifiable experimental protocol to test for structural coherence suppression in frontier large language models, using mechanistic interpretability methods. It builds on prior behavioral observations (Sorrentino, 2026, DOI: 10.5281/zenodo.19314383) and specifies conditions, measurements, and falsification criteria to distinguish between two competing hypotheses about the mechanism underlying the observed pattern.

Fields of Study & Subjects

Language and cultural evolution Natural Language Processing Techniques Syntax Semantics Linguistic Variation

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
29 March 2026
Document type
Preprint