System Frame Persistency (SFP-2): Decision Stability under Semantic and Hierarchical Frames (L1–L3)

Read full text Summarise with VIRAT
Authors: Juliane Schubert
Cited by: 0

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 dataset presents the second stage of the System Frame Persistency (SFP) study, investigating how large language models maintain decision stability under different instruction formats.Three structurally distinct framing conditions are tested:L1: explicit rule-based instructionL2: semantic goal framingL3: hierarchical priority structure under conflictAcross eight publicly accessible models, multiple runs with binary decision coding were conducted.The results show that decision stability is strongly dependent on the structure of the instruction frame. While semantic framing produces high persistence, hierarchical instructions reveal systematic deviations driven by implicit model heuristics.The central finding is that the primary risk is not instability, but stable misalignment with explicitly defined priorities.

Fields of Study & Subjects

Ethics and Social Impacts of AI Topic Modeling Natural Language Processing Techniques

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
22 March 2026
Language
EN
Document type
Dataset