Data and Code for "Do Neuromodulator-Inspired State Prompts Change LLM Behavior?"

This record contains the data and code accompanying the manuscript “Do Neuromodulator-Inspired State Prompts Change LLM Behavior? A Locally Preregistered Controlled Audit of State Labels, Behavioral Consequences, and Direct Instructions.” The release includes the complete portable Study 1 archive and a selected sealed evidence bundle for the component-matched Study 2 replication and its GLM-5.2 semantic audit. Materials include protocols, SHA-256 manifests, immutable result reports, raw Study 1 responses and execution records, condition-blind human adjudications, deterministic analysis programs, a claim-evidence matrix, and privacy-redaction documentation. The studies evaluate causal effects of fixed natural-language prompt packages on observable outputs from one tested language-model configuration. The materials do not support claims about subjective fatigue, curiosity, consciousness, biological neuromodulation, persistent internal state, or a unique internal mechanism. Study 2 full target-call and judge-call execution trees are retained in the sealed formal archive but are not duplicated in this compact public release. The included README describes the evidence hierarchy, artifact map, reproduction scope, and verification commands.