Uncovering the Internal Effectiveness of Static LLM Lineage Construction

Read full text Summarise with VIRAT
Authors: Anonymous Authors
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

Abstract Large Language Models (LLMs) are increasingly reused and adapted using fine-tuning, merging, and quantization, enabling customization and efficiency improvements. Ensuring transparency in model composition is critical for risk mitigation. However, existing mechanisms like model cards are incompletely or incorrectly disclosed. In this work, we focus on model lineage construction for full-parameter fine-tuning via static fingerprints to trace the derivation and relationships among LLMs. Through an empirical study, we systematically investigate key factors affecting lineage construction, including construction dimensions, failure patterns and the effects of model-component selection. Based on the revealed findings, we further investigate two heuristics to improve LLM lineage construction. Our findings demonstrate that optimized strategies can enhance lineage-construction performance. This paper has been submitted to FSE 2027.

Fields of Study & Subjects

Software Engineering Research Scientific Computing and Data Management Explainable Artificial Intelligence (XAI)

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
27 March 2026
Language
EN