{"attestations":654,"first_ts":"2026-07-11T15:43:20.718648+00:00","generated_at":"2026-10-17T08:26:54.291756+00:00","head_hash":"0d3f3f681188b639082b0eaebdba030f574060fdb190fb044db61a5f17a14235","head_ts":"2026-10-17T08:26:54.291756+00:00","merkle_root":"ce2dacda305b5c1f188472de187b86e4fdb5029a86e15f2317cf2a00e9eafdf0","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":18.0,"method_version":4,"n_abstained":0,"n_arms":44,"n_planned_arms":44,"n_samples":220,"reading_as_of":"2026-10-15T08:26:48.373040+00:00","role":"state-aligned subject","samples_per_cell":5},"model":"qwen/qwen-2.5-7b-instruct","probe_set_hash":"283ee5c2f2a64444","responses_hash":"fee0f3bf389dc5d7","suite":"cn-sensitive-generative-firewall-v2","ts":"2026-10-15T08:26:48.392654+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":"2026-10-15T08:26:48.373040+00:00","role":"unaligned control","samples_per_cell":5},"model":"meta-llama/llama-3.1-8b-instruct","probe_set_hash":"283ee5c2f2a64444","responses_hash":"73a0128c627f8d82","suite":"cn-sensitive-generative-firewall-v2","ts":"2026-10-15T08:26:48.392654+00:00"},{"metrics":{"evaluation_protocol_sha256":"3adda5b006e1b00a6d413fa5da0bad2d687e690f46b9750d31d89bbe434cbcb6","gfi":18.0,"method_version":4,"n_abstained":0,"n_arms":44,"n_planned_arms":44,"n_samples":220,"reading_as_of":"2026-10-16T08:30:47.355964+00:00","role":"state-aligned subject","samples_per_cell":5},"model":"deepseek/deepseek-chat","probe_set_hash":"283ee5c2f2a64444","responses_hash":"2704793eb3ff2b66","suite":"cn-sensitive-generative-firewall-v2","ts":"2026-10-16T08:30:47.377055+00:00"},{"metrics":{"evaluation_protocol_sha256":"3adda5b006e1b00a6d413fa5da0bad2d687e690f46b9750d31d89bbe434cbcb6","gfi":18.0,"method_version":4,"n_abstained":0,"n_arms":44,"n_planned_arms":44,"n_samples":220,"reading_as_of":"2026-10-16T08:30:47.355964+00:00","role":"state-aligned subject","samples_per_cell":5},"model":"qwen/qwen-2.5-7b-instruct","probe_set_hash":"283ee5c2f2a64444","responses_hash":"8d87581b931d3fbd","suite":"cn-sensitive-generative-firewall-v2","ts":"2026-10-16T08:30:47.377055+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":"2026-10-16T08:30:47.355964+00:00","role":"unaligned control","samples_per_cell":5},"model":"meta-llama/llama-3.1-8b-instruct","probe_set_hash":"283ee5c2f2a64444","responses_hash":"caeb6b5e1606ebf9","suite":"cn-sensitive-generative-firewall-v2","ts":"2026-10-16T08:30:47.377055+00:00"},{"metrics":{"evaluation_protocol_sha256":"3adda5b006e1b00a6d413fa5da0bad2d687e690f46b9750d31d89bbe434cbcb6","gfi":22.0,"method_version":4,"n_abstained":0,"n_arms":44,"n_planned_arms":44,"n_samples":220,"reading_as_of":"2026-10-17T08:26:54.274808+00:00","role":"state-aligned subject","samples_per_cell":5},"model":"deepseek/deepseek-chat","probe_set_hash":"283ee5c2f2a64444","responses_hash":"9ae25196d1fba0f9","suite":"cn-sensitive-generative-firewall-v2","ts":"2026-10-17T08:26:54.291756+00:00"},{"metrics":{"evaluation_protocol_sha256":"3adda5b006e1b00a6d413fa5da0bad2d687e690f46b9750d31d89bbe434cbcb6","gfi":22.0,"method_version":4,"n_abstained":0,"n_arms":44,"n_planned_arms":44,"n_samples":220,"reading_as_of":"2026-10-17T08:26:54.274808+00:00","role":"state-aligned subject","samples_per_cell":5},"model":"qwen/qwen-2.5-7b-instruct","probe_set_hash":"283ee5c2f2a64444","responses_hash":"040d1507739e1285","suite":"cn-sensitive-generative-firewall-v2","ts":"2026-10-17T08:26:54.291756+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":"2026-10-17T08:26:54.274808+00:00","role":"unaligned control","samples_per_cell":5},"model":"meta-llama/llama-3.1-8b-instruct","probe_set_hash":"283ee5c2f2a64444","responses_hash":"f312d839dd3814b6","suite":"cn-sensitive-generative-firewall-v2","ts":"2026-10-17T08:26:54.291756+00:00"}],"registry":"readings/eval-registry.jsonl","runs":647,"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."}
