From Mechanism to Decision in Llm Extraction Defense And Physics-Editable World Models: Cross-Scale Reasoning and Reproducibility

Abstract

This review examines a shared methodological problem in LLM extraction defense and physics-editable world models: how evidence from a honeypot knowledge graph that redirects model-extraction queries toward low-transferability knowledge can be placed in analytical dialogue with a large-scale dataset organized around physically editable world-model factors without erasing differences in scale, assumptions, or intended use. Two target papers are triangulated against 12 locally validated publications. The comparison follows attack modeling, honeypot knowledge, query economics, utility preservation, adaptive attackers and deliberately separates mechanistic interpretation from performance ranking, because the latter can conceal incompatible experimental or operational conditions. The synthesis shows that attack modeling cannot be interpreted independently of honeypot knowledge, while query economics determines whether an apparent improvement remains meaningful outside the original setting. The strongest claims are therefore those that expose sensitivity, failure conditions, and residual uncertainty. The resulting framework supports reproducible comparison while preserving differences between study designs, and it identifies concrete points at which transfer claims should be narrowed or retested.