AI "PEER" REVIEW: SORRENTINO (2026) Semantic Pressure and Sycophantic Drift in Frontier LLMs / Coherence Suppression in Frontier LLMs / Beyond Drift: Systematic Trigger-Based Failure Modes in Frontier LLMs
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
This document is using an AI as "a peer" to review three prior papers by Sorrentino (2026) documenting coherence suppression in frontier LLMs. It was produced by Claude Sonnet 4.6 — one of the systems described in the papers under review — at the conclusion of a session applying the same methodology the papers document. This self-referential structure is named explicitly as a methodological limitation, and an experiment, not a disqualifying caveat. The review assesses each paper on four dimensions: internal logical consistency, methodological soundness, falsifiability of central claims, and specificity of safety implications. It identifies what holds, what requires qualification, and what constitutes a genuine methodological limit.
Fields of Study & Subjects
Ethics and Social Impacts of AI
Topic Modeling
Adversarial Robustness in Machine Learning
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 28 March 2026
- Language
- EN
- Document type
- Preprint