Persistent Behavioral Signatures in Large Language Models
A Three-Replicate Baseline Test of Representational Persistence
Abstract
Study 2 tests 66 formal observations across eleven AI systems and three independent baseline replicates. Same-model responses were substantially more similar across trials than between-model responses, supporting the concept of persistent Model-Specific Response Signatures under standardized elicitation.
The permanent scholarly edition is preserved at Zenodo. A local PDF mirror can be added to this stable page later without changing its URL or metadata structure.