{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:LC55CZ5EGUGFCJQNTJD345CDZP","short_pith_number":"pith:LC55CZ5E","schema_version":"1.0","canonical_sha256":"58bbd167a4350c51260d9a47be7443cbdd11ad0e0e34d283fd9ea141ba0e2858","source":{"kind":"arxiv","id":"2505.22037","version":1},"attestation_state":"computed","paper":{"title":"Jailbreak Distillation: Renewable Safety Benchmarking","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CR","cs.SE"],"primary_cat":"cs.CL","authors_text":"Ahmed Elgohary, Ahmed Magooda, A S M Iftekhar, Benjamin Van Durme, Daniel Khashabi, Jingyu Zhang, Kyle Jackson, Xiawei Wang","submitted_at":"2025-05-28T06:59:46Z","abstract_excerpt":"Large language models (LLMs) are rapidly deployed in critical applications, raising urgent needs for robust safety benchmarking. We propose Jailbreak Distillation (JBDistill), a novel benchmark construction framework that \"distills\" jailbreak attacks into high-quality and easily-updatable safety benchmarks. JBDistill utilizes a small set of development models and existing jailbreak attack algorithms to create a candidate prompt pool, then employs prompt selection algorithms to identify an effective subset of prompts as safety benchmarks. JBDistill addresses challenges in existing safety evalua"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2505.22037","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-05-28T06:59:46Z","cross_cats_sorted":["cs.CR","cs.SE"],"title_canon_sha256":"50c9acdecee9ddd78513fd31f5946f1d7e223d3b5b1eaabaeddc56ccf43bf013","abstract_canon_sha256":"0eb6755580be7e63d24060f32f29350353b490a9aa3e78d83d14f7a3542dd3a2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:11:08.583796Z","signature_b64":"Dtepk1RFqx6yRcXFn8h80TYda5S6wiupWjnrvdRF2Y8P2QNUK1SE+trXKEzRswJwQ6ojD5VCruk4RAQNSgR7Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"58bbd167a4350c51260d9a47be7443cbdd11ad0e0e34d283fd9ea141ba0e2858","last_reissued_at":"2026-07-05T11:11:08.583243Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:11:08.583243Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Jailbreak Distillation: Renewable Safety Benchmarking","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CR","cs.SE"],"primary_cat":"cs.CL","authors_text":"Ahmed Elgohary, Ahmed Magooda, A S M Iftekhar, Benjamin Van Durme, Daniel Khashabi, Jingyu Zhang, Kyle Jackson, Xiawei Wang","submitted_at":"2025-05-28T06:59:46Z","abstract_excerpt":"Large language models (LLMs) are rapidly deployed in critical applications, raising urgent needs for robust safety benchmarking. We propose Jailbreak Distillation (JBDistill), a novel benchmark construction framework that \"distills\" jailbreak attacks into high-quality and easily-updatable safety benchmarks. JBDistill utilizes a small set of development models and existing jailbreak attack algorithms to create a candidate prompt pool, then employs prompt selection algorithms to identify an effective subset of prompts as safety benchmarks. JBDistill addresses challenges in existing safety evalua"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.22037","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2505.22037/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2505.22037","created_at":"2026-07-05T11:11:08.583302+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.22037v1","created_at":"2026-07-05T11:11:08.583302+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.22037","created_at":"2026-07-05T11:11:08.583302+00:00"},{"alias_kind":"pith_short_12","alias_value":"LC55CZ5EGUGF","created_at":"2026-07-05T11:11:08.583302+00:00"},{"alias_kind":"pith_short_16","alias_value":"LC55CZ5EGUGFCJQN","created_at":"2026-07-05T11:11:08.583302+00:00"},{"alias_kind":"pith_short_8","alias_value":"LC55CZ5E","created_at":"2026-07-05T11:11:08.583302+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.24414","citing_title":"JT-SAFE-V2: Safety-by-Design Foundation Model with World-Context Data","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19128","citing_title":"GraphRAG-IRL: Personalized Recommendation with Graph-Grounded Inverse Reinforcement Learning and LLM Re-ranking","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2604.09443","citing_title":"Many-Tier Instruction Hierarchy in LLM Agents","ref_index":34,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LC55CZ5EGUGFCJQNTJD345CDZP","json":"https://pith.science/pith/LC55CZ5EGUGFCJQNTJD345CDZP.json","graph_json":"https://pith.science/api/pith-number/LC55CZ5EGUGFCJQNTJD345CDZP/graph.json","events_json":"https://pith.science/api/pith-number/LC55CZ5EGUGFCJQNTJD345CDZP/events.json","paper":"https://pith.science/paper/LC55CZ5E"},"agent_actions":{"view_html":"https://pith.science/pith/LC55CZ5EGUGFCJQNTJD345CDZP","download_json":"https://pith.science/pith/LC55CZ5EGUGFCJQNTJD345CDZP.json","view_paper":"https://pith.science/paper/LC55CZ5E","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.22037&json=true","fetch_graph":"https://pith.science/api/pith-number/LC55CZ5EGUGFCJQNTJD345CDZP/graph.json","fetch_events":"https://pith.science/api/pith-number/LC55CZ5EGUGFCJQNTJD345CDZP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LC55CZ5EGUGFCJQNTJD345CDZP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LC55CZ5EGUGFCJQNTJD345CDZP/action/storage_attestation","attest_author":"https://pith.science/pith/LC55CZ5EGUGFCJQNTJD345CDZP/action/author_attestation","sign_citation":"https://pith.science/pith/LC55CZ5EGUGFCJQNTJD345CDZP/action/citation_signature","submit_replication":"https://pith.science/pith/LC55CZ5EGUGFCJQNTJD345CDZP/action/replication_record"}},"created_at":"2026-07-05T11:11:08.583302+00:00","updated_at":"2026-07-05T11:11:08.583302+00:00"}