When stream averages suffice: a protocol for evaluating adaptive neural dynamics (research artifact, R6)

Research artifact (release R6) for the paper "Evaluating Adaptation Dynamics in Neural Systems: When Are Short-Stream Averages Sufficient?" (submitted to Neurocomputing). The archive contains the perturbation generator, domain adapters (CIFAR-10, CIFAR-100, ETTh1, ETTh2), sixteen adaptation-policy implementations, frozen registered configurations, per-stream traces, integrity ledgers, the interval-coverage study, and every analysis used in the paper, bound to immutable content by a SHA-256 source manifest and result ledgers. Verification entry points: check_release.sh (manifest, tests, authoritative results) and run_smoke.sh. Archive SHA-256: 03ea5b9ee87f4a6b364a6d46dd9c5d429dfcc964b2d102ff79a5f53ade7dde57. License: MIT. Python 3.