Release materials for "Satisfied, Not Solving: Wireheading of Bounded Homeostatic Rewards in Reinforcement Learning"

Release materials for the preprint "Satisfied, Not Solving: Wireheading of Bounded Homeostatic Rewards in Reinforcement Learning" (C. J. Metoyer). Contains the pre-registration files and their sha256 hash records (rounds 3–5), run records for all 57 training runs (configs, logs, per-seed metrics), the scored result files, final policy checkpoints, the training code as used in the paper (PPO and the homeostatic satiation reward), and the paper's data, figure, and analysis scripts. Lock times in the hash records are self-reported. README.md maps every file to the paper section that uses it. Version 1.0.1 is a snapshot of the public repository github.com/chasemetoyer/satisfied-not-solving (tag v1.0.1): code, data, every result file, and a Makefile that regenerates them. Version 1.0 holds the original release materials