{"schema_version":"papers/paper-detail-v1","title":"Inverting the Shield: Systematically Generating Safety Tests from Policy Specifications","surface":"papers","opportunity_kernel":{"paper_id":"45801f8b-cd6f-4961-8a7b-ae1d694716b5","title":"Inverting the Shield: Systematically Generating Safety Tests from Policy Specifications","authors":["Xiaoyue Lu","Xianglin Yang","Haijun Liu","Jiahao Liu","Kuntai Cai","Yan Xiao","Jin Song Dong"],"arxiv_id":"2605.24883v1","doi":null,"published_at":"2026-05-24T05:51:23.000Z","score_object":{"overall":{"value":0,"scale":"0-10","confidence":0.85,"confidence_reason":"Backfilled from persisted papers.viability_score.","model_version":"phase0-backfill-v1","pipeline_version":"phase0-kernel-v1","computed_at":"2026-05-27T01:01:19.691Z","fresh_until":"2026-06-26T01:01:19.691Z","is_stale":false,"source_count":1,"missingness":[]},"technical":{"value":1.4,"scale":"0-10","confidence":0.55,"confidence_reason":"Backfilled from paper_extraction_scorecards.reconstruction_score.","model_version":"phase0-backfill-v1","pipeline_version":"phase0-kernel-v1","computed_at":"2026-05-27T01:01:55.330Z","fresh_until":"2026-06-10T01:01:55.330Z","is_stale":false,"source_count":1,"missingness":["reproducibility_results.reproducibility_score","deployability_scores.score"]},"commercial":{"value":5,"scale":"0-10","confidence":0.75,"confidence_reason":"Backfilled from persisted commercial_flags and repo availability.","model_version":"phase0-backfill-v1","pipeline_version":"phase0-kernel-v1","computed_at":"2026-05-27T01:01:19.691Z","fresh_until":"2026-06-26T01:01:19.691Z","is_stale":false,"source_count":2,"missingness":[]},"market":{"value":5.5,"scale":"0-10","confidence":0.45,"confidence_reason":"Heuristic market score materialized from paper metadata, repo availability, and deployability.","model_version":"phase0-backfill-v1","pipeline_version":"phase0-kernel-v1","computed_at":"2026-05-27T01:10:35.163Z","fresh_until":"2026-06-10T01:10:35.163Z","is_stale":false,"source_count":1,"missingness":[]},"team":{"value":4.7,"scale":"0-10","confidence":0.28,"confidence_reason":"Heuristic fallback from paper author count.","model_version":"phase0-backfill-v1","pipeline_version":"phase0-kernel-v1","computed_at":"2026-05-27T01:01:19.691Z","fresh_until":"2026-06-10T01:01:19.691Z","is_stale":false,"source_count":1,"missingness":["paper_author_affiliations"]},"methodology":{"value":3.9,"scale":"0-10","confidence":0.82,"confidence_reason":"Backfilled from paper_extraction_scorecards.total_score.","model_version":"phase0-backfill-v1","pipeline_version":"phase0-kernel-v1","computed_at":"2026-05-27T01:01:55.330Z","fresh_until":"2026-06-26T01:01:55.330Z","is_stale":false,"source_count":2,"missingness":[]}},"evidence_receipt":{"freshness":"fresh","proof_status":"unverified","repo_status":"active","references_count":0,"source_count":4,"coverage":0.6667,"missingness":["references","proof_status"],"unresolved_unknowns":["proof verification has not been recorded yet"],"last_verification_at":"2026-05-27T01:10:35.163Z"},"lineage_hash":"7a815bede37c2a3a9362cb15fa4dcec7c884a4b109b5887c6508b36c0924717c"},"distribution":null,"replication_evidence":[],"author_dna":[]}