Data and Results for: When Forgetting Fails: Characterizing Class Unlearning Failure Modes in Non-IID Decentralized Federated Learning

Experimental results, model weights, and training logs for the paper "When Forgetting Fails: Characterizing Class Unlearning Failure Modes in Non-IID Decentralized Federated Learning". Contains results from a physical six-node Raspberry Pi Zero 2W cluster (primary experiments) and a 30-node Mac validation experiment. Includes per-node model weights, training logs, unlearning outputs, and summary CSVs mapping to paper tables. See README.md for full directory structure and mapping to paper tables.