Evidence-bound AI systems: agent-benchmark audits, tool effects, checkpoint restart invariants, and fail-closed recovery.
rust open-source distributed-systems pytorch reproducibility fault-injection research-engineering ai-systems agent-evaluation agent-harness benchmark-integrity inference-systems systems-testing
-
Updated
Jul 29, 2026