Christopher Altman

Reproducible evaluation harnesses for frontier AI and quantum systems

Building reproducible instruments for frontier AI evaluation, alignment testing, and quantum-system benchmarks.

Our research program develops reproducible evaluation harnesses for testing claims in frontier AI, quantum machine learning, and superconducting quantum systems. We combine quantum machine learning methods with classical testbeds to probe function–representation decoupling, continuation interests, and the reliability of geometric proxies in model repair.

Current work spans structural AI evaluation, continuation-risk measurement, quantum kernel methods for telemetry anomaly detection, binarized quantum neural networks, Hamiltonian dynamics for data augmentation, recursive self-improvement, satellite quantum key distribution, superconducting-qubit simulation, and hardware benchmarks for Wigner’s-friend-style circuits. Each project is packaged as a reproducible evaluation harness with diagnostic plots, baselines, and reviewable artifacts designed to test specific technical claims under controlled experimental conditions.