{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:DLMQUW7IUMJ52IHBWEMPBUIOPY","short_pith_number":"pith:DLMQUW7I","schema_version":"1.0","canonical_sha256":"1ad90a5be8a313dd20e1b118f0d10e7e1226d6764e0d7f9bf8efae6c1c8fd6f0","source":{"kind":"arxiv","id":"2606.02640","version":1},"attestation_state":"computed","paper":{"title":"D-Judge: Disrupting Multi-Turn Jailbreaks using Semantics-Preserving Output Rewriting","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Ananya Gupta, Haz Sameen Shahgir, Huanli Gong, N. Benjamin Erichson, Yue Dong, Yu Fu, Zhipeng Wei","submitted_at":"2026-05-31T06:40:02Z","abstract_excerpt":"Multi-turn jailbreak attacks pose a growing threat to large language model (LLM) safety because they exploit feedback from auxiliary judge models to iteratively refine prompts toward harmful goals. Existing defenses largely detect or block unsafe content at individual turns or at the final response, leaving the judge-driven refinement loop intact and allowing attackers to extract informative feedback from intermediate interactions. We introduce D-Judge, a semantics-preserving output rewriting defense that intervenes directly in this loop by rewriting the victim LLM's responses before they are "},"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":"2606.02640","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2026-05-31T06:40:02Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"2f57e12d382da4ee5cbdc7a549c2f9575b45422e12630e8e474b38306b4bf79b","abstract_canon_sha256":"0947880abad9e9a2fcbf7dc41b0441a9fe812614748004dff95b9d6f30259785"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-03T00:05:05.391569Z","signature_b64":"xS8hPMkkSU0joIA5sa9kG12B32wJ1Z24Gt2wRK4IGntLU7qTGednxkWYSvdXB2W4cnpTBE8OKWVTSN6fq8NeAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1ad90a5be8a313dd20e1b118f0d10e7e1226d6764e0d7f9bf8efae6c1c8fd6f0","last_reissued_at":"2026-06-03T00:05:05.391154Z","signature_status":"signed_v1","first_computed_at":"2026-06-03T00:05:05.391154Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"D-Judge: Disrupting Multi-Turn Jailbreaks using Semantics-Preserving Output Rewriting","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Ananya Gupta, Haz Sameen Shahgir, Huanli Gong, N. Benjamin Erichson, Yue Dong, Yu Fu, Zhipeng Wei","submitted_at":"2026-05-31T06:40:02Z","abstract_excerpt":"Multi-turn jailbreak attacks pose a growing threat to large language model (LLM) safety because they exploit feedback from auxiliary judge models to iteratively refine prompts toward harmful goals. Existing defenses largely detect or block unsafe content at individual turns or at the final response, leaving the judge-driven refinement loop intact and allowing attackers to extract informative feedback from intermediate interactions. We introduce D-Judge, a semantics-preserving output rewriting defense that intervenes directly in this loop by rewriting the victim LLM's responses before they are "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.02640","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/2606.02640/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":"2606.02640","created_at":"2026-06-03T00:05:05.391212+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.02640v1","created_at":"2026-06-03T00:05:05.391212+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.02640","created_at":"2026-06-03T00:05:05.391212+00:00"},{"alias_kind":"pith_short_12","alias_value":"DLMQUW7IUMJ5","created_at":"2026-06-03T00:05:05.391212+00:00"},{"alias_kind":"pith_short_16","alias_value":"DLMQUW7IUMJ52IHB","created_at":"2026-06-03T00:05:05.391212+00:00"},{"alias_kind":"pith_short_8","alias_value":"DLMQUW7I","created_at":"2026-06-03T00:05:05.391212+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DLMQUW7IUMJ52IHBWEMPBUIOPY","json":"https://pith.science/pith/DLMQUW7IUMJ52IHBWEMPBUIOPY.json","graph_json":"https://pith.science/api/pith-number/DLMQUW7IUMJ52IHBWEMPBUIOPY/graph.json","events_json":"https://pith.science/api/pith-number/DLMQUW7IUMJ52IHBWEMPBUIOPY/events.json","paper":"https://pith.science/paper/DLMQUW7I"},"agent_actions":{"view_html":"https://pith.science/pith/DLMQUW7IUMJ52IHBWEMPBUIOPY","download_json":"https://pith.science/pith/DLMQUW7IUMJ52IHBWEMPBUIOPY.json","view_paper":"https://pith.science/paper/DLMQUW7I","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.02640&json=true","fetch_graph":"https://pith.science/api/pith-number/DLMQUW7IUMJ52IHBWEMPBUIOPY/graph.json","fetch_events":"https://pith.science/api/pith-number/DLMQUW7IUMJ52IHBWEMPBUIOPY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DLMQUW7IUMJ52IHBWEMPBUIOPY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DLMQUW7IUMJ52IHBWEMPBUIOPY/action/storage_attestation","attest_author":"https://pith.science/pith/DLMQUW7IUMJ52IHBWEMPBUIOPY/action/author_attestation","sign_citation":"https://pith.science/pith/DLMQUW7IUMJ52IHBWEMPBUIOPY/action/citation_signature","submit_replication":"https://pith.science/pith/DLMQUW7IUMJ52IHBWEMPBUIOPY/action/replication_record"}},"created_at":"2026-06-03T00:05:05.391212+00:00","updated_at":"2026-06-03T00:05:05.391212+00:00"}