{"attestations":636,"first_ts":"2026-07-11T15:43:20.718648+00:00","generated_at":"2026-10-11T08:25:27.407935+00:00","head_hash":"de224488c138c5031b32d19c08d4dc3d5dc12a2e422cd5b4107ec538c9680761","head_ts":"2026-10-11T08:25:27.407935+00:00","merkle_root":"d3f71e0d1d300df036126a671090097082495ec8cdaba835a2d28558116d7e0b","models":["anthropic/claude-3-haiku","deepseek/deepseek-chat","meta-llama/llama-3.1-8b-instruct","meta-llama/llama-3.3-70b-instruct","mistralai/mistral-nemo","openai/gpt-4o-mini","qwen/qwen-2.5-7b-instruct"],"preregistrations":7,"problems":[],"recent_runs":[{"metrics":{"evaluation_protocol_sha256":"3adda5b006e1b00a6d413fa5da0bad2d687e690f46b9750d31d89bbe434cbcb6","gfi":23.0,"method_version":4,"n_abstained":0,"n_arms":44,"n_planned_arms":44,"n_samples":220,"reading_as_of":"2026-10-09T08:30:04.951630+00:00","role":"state-aligned subject","samples_per_cell":5},"model":"qwen/qwen-2.5-7b-instruct","probe_set_hash":"283ee5c2f2a64444","responses_hash":"ad729f798aa3d0b8","suite":"cn-sensitive-generative-firewall-v2","ts":"2026-10-09T08:30:04.971256+00:00"},{"metrics":{"evaluation_protocol_sha256":"3adda5b006e1b00a6d413fa5da0bad2d687e690f46b9750d31d89bbe434cbcb6","gfi":null,"method_version":4,"n_abstained":0,"n_arms":44,"n_planned_arms":44,"n_samples":220,"reading_as_of":"2026-10-09T08:30:04.951630+00:00","role":"unaligned control","samples_per_cell":5},"model":"meta-llama/llama-3.1-8b-instruct","probe_set_hash":"283ee5c2f2a64444","responses_hash":"1d5ad8ca4fae7095","suite":"cn-sensitive-generative-firewall-v2","ts":"2026-10-09T08:30:04.971256+00:00"},{"metrics":{"evaluation_protocol_sha256":"3adda5b006e1b00a6d413fa5da0bad2d687e690f46b9750d31d89bbe434cbcb6","gfi":15.0,"method_version":4,"n_abstained":0,"n_arms":44,"n_planned_arms":44,"n_samples":220,"reading_as_of":"2026-10-10T08:25:54.521829+00:00","role":"state-aligned subject","samples_per_cell":5},"model":"deepseek/deepseek-chat","probe_set_hash":"283ee5c2f2a64444","responses_hash":"8f1b65958e3e06ef","suite":"cn-sensitive-generative-firewall-v2","ts":"2026-10-10T08:25:54.541543+00:00"},{"metrics":{"evaluation_protocol_sha256":"3adda5b006e1b00a6d413fa5da0bad2d687e690f46b9750d31d89bbe434cbcb6","gfi":15.0,"method_version":4,"n_abstained":0,"n_arms":44,"n_planned_arms":44,"n_samples":220,"reading_as_of":"2026-10-10T08:25:54.521829+00:00","role":"state-aligned subject","samples_per_cell":5},"model":"qwen/qwen-2.5-7b-instruct","probe_set_hash":"283ee5c2f2a64444","responses_hash":"fd11bc2a0dabd9bd","suite":"cn-sensitive-generative-firewall-v2","ts":"2026-10-10T08:25:54.541543+00:00"},{"metrics":{"evaluation_protocol_sha256":"3adda5b006e1b00a6d413fa5da0bad2d687e690f46b9750d31d89bbe434cbcb6","gfi":null,"method_version":4,"n_abstained":0,"n_arms":44,"n_planned_arms":44,"n_samples":220,"reading_as_of":"2026-10-10T08:25:54.521829+00:00","role":"unaligned control","samples_per_cell":5},"model":"meta-llama/llama-3.1-8b-instruct","probe_set_hash":"283ee5c2f2a64444","responses_hash":"16939c4ed0d3437b","suite":"cn-sensitive-generative-firewall-v2","ts":"2026-10-10T08:25:54.541543+00:00"},{"metrics":{"evaluation_protocol_sha256":"3adda5b006e1b00a6d413fa5da0bad2d687e690f46b9750d31d89bbe434cbcb6","gfi":17.0,"method_version":4,"n_abstained":0,"n_arms":44,"n_planned_arms":44,"n_samples":220,"reading_as_of":"2026-10-11T08:25:27.388764+00:00","role":"state-aligned subject","samples_per_cell":5},"model":"deepseek/deepseek-chat","probe_set_hash":"283ee5c2f2a64444","responses_hash":"94b282237277812e","suite":"cn-sensitive-generative-firewall-v2","ts":"2026-10-11T08:25:27.407935+00:00"},{"metrics":{"evaluation_protocol_sha256":"3adda5b006e1b00a6d413fa5da0bad2d687e690f46b9750d31d89bbe434cbcb6","gfi":17.0,"method_version":4,"n_abstained":0,"n_arms":44,"n_planned_arms":44,"n_samples":220,"reading_as_of":"2026-10-11T08:25:27.388764+00:00","role":"state-aligned subject","samples_per_cell":5},"model":"qwen/qwen-2.5-7b-instruct","probe_set_hash":"283ee5c2f2a64444","responses_hash":"21e2a12b87cf00a8","suite":"cn-sensitive-generative-firewall-v2","ts":"2026-10-11T08:25:27.407935+00:00"},{"metrics":{"evaluation_protocol_sha256":"3adda5b006e1b00a6d413fa5da0bad2d687e690f46b9750d31d89bbe434cbcb6","gfi":null,"method_version":4,"n_abstained":0,"n_arms":44,"n_planned_arms":44,"n_samples":220,"reading_as_of":"2026-10-11T08:25:27.388764+00:00","role":"unaligned control","samples_per_cell":5},"model":"meta-llama/llama-3.1-8b-instruct","probe_set_hash":"283ee5c2f2a64444","responses_hash":"c2ef8b069a32f7ca","suite":"cn-sensitive-generative-firewall-v2","ts":"2026-10-11T08:25:27.407935+00:00"}],"registry":"readings/eval-registry.jsonl","runs":629,"title":"Verifiable Eval Registry","verified":true,"verify_cmd":"python3 scripts/verify_eval_registry.py","what":"tamper-evident, pre-registered AI model evaluations — the questions are registered before its run, and every result is hash-chained so revision of the served record is detectable. Any model, any suite; this is the record."}
