{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:EQKC263GSRT36XHXXNMVRAAJE5","short_pith_number":"pith:EQKC263G","schema_version":"1.0","canonical_sha256":"24142d7b669467bf5cf7bb59588009277b69f6f9b7e70a5310cdb4b34fc0e5b3","source":{"kind":"arxiv","id":"2606.25059","version":1},"attestation_state":"computed","paper":{"title":"What Does It Mean to Break a Distillation Defense?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Daniel Paleka, Florian Tram\\`er, Lena Libon, Michael Aerni, Pura Peetathawatchai","submitted_at":"2026-06-23T18:13:02Z","abstract_excerpt":"Black-box LLMs (accessible only via API) are vulnerable to distillation attacks, in which an attacker queries the model and trains a student on its outputs. A recent line of work proposes output perturbation defenses that modify the teacher's output to reduce student performance while preserving utility for legitimate users. As a relatively new family of approaches, output perturbation defenses lack a shared threat model, making it difficult to compare them, reason about composing them with other attacks, or evaluate their robustness against realistic adversaries. This underspecification matte"},"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":"2606.25059","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2026-06-23T18:13:02Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"2931b6ae396b0f678776891bb5c3ccbc9f90b6a261165481e62dc02103d24654","abstract_canon_sha256":"4a0a02228d204b7b0ffdfd72f9acf0eab286e0f970dd27630f12c17d7c67a52f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-25T00:18:16.635574Z","signature_b64":"AwiXuf22gvqx28Rq34Zy2vxbWG23NT0Xsu/8AsC6qABcmPymxc9/SV8ZOeFRQkUCLRzP6FEnCPlLEUOelWW3BA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"24142d7b669467bf5cf7bb59588009277b69f6f9b7e70a5310cdb4b34fc0e5b3","last_reissued_at":"2026-06-25T00:18:16.635189Z","signature_status":"signed_v1","first_computed_at":"2026-06-25T00:18:16.635189Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"What Does It Mean to Break a Distillation Defense?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Daniel Paleka, Florian Tram\\`er, Lena Libon, Michael Aerni, Pura Peetathawatchai","submitted_at":"2026-06-23T18:13:02Z","abstract_excerpt":"Black-box LLMs (accessible only via API) are vulnerable to distillation attacks, in which an attacker queries the model and trains a student on its outputs. A recent line of work proposes output perturbation defenses that modify the teacher's output to reduce student performance while preserving utility for legitimate users. As a relatively new family of approaches, output perturbation defenses lack a shared threat model, making it difficult to compare them, reason about composing them with other attacks, or evaluate their robustness against realistic adversaries. This underspecification matte"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.25059","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/2606.25059/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":"2606.25059","created_at":"2026-06-25T00:18:16.635246+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.25059v1","created_at":"2026-06-25T00:18:16.635246+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.25059","created_at":"2026-06-25T00:18:16.635246+00:00"},{"alias_kind":"pith_short_12","alias_value":"EQKC263GSRT3","created_at":"2026-06-25T00:18:16.635246+00:00"},{"alias_kind":"pith_short_16","alias_value":"EQKC263GSRT36XHX","created_at":"2026-06-25T00:18:16.635246+00:00"},{"alias_kind":"pith_short_8","alias_value":"EQKC263G","created_at":"2026-06-25T00:18:16.635246+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/EQKC263GSRT36XHXXNMVRAAJE5","json":"https://pith.science/pith/EQKC263GSRT36XHXXNMVRAAJE5.json","graph_json":"https://pith.science/api/pith-number/EQKC263GSRT36XHXXNMVRAAJE5/graph.json","events_json":"https://pith.science/api/pith-number/EQKC263GSRT36XHXXNMVRAAJE5/events.json","paper":"https://pith.science/paper/EQKC263G"},"agent_actions":{"view_html":"https://pith.science/pith/EQKC263GSRT36XHXXNMVRAAJE5","download_json":"https://pith.science/pith/EQKC263GSRT36XHXXNMVRAAJE5.json","view_paper":"https://pith.science/paper/EQKC263G","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.25059&json=true","fetch_graph":"https://pith.science/api/pith-number/EQKC263GSRT36XHXXNMVRAAJE5/graph.json","fetch_events":"https://pith.science/api/pith-number/EQKC263GSRT36XHXXNMVRAAJE5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EQKC263GSRT36XHXXNMVRAAJE5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EQKC263GSRT36XHXXNMVRAAJE5/action/storage_attestation","attest_author":"https://pith.science/pith/EQKC263GSRT36XHXXNMVRAAJE5/action/author_attestation","sign_citation":"https://pith.science/pith/EQKC263GSRT36XHXXNMVRAAJE5/action/citation_signature","submit_replication":"https://pith.science/pith/EQKC263GSRT36XHXXNMVRAAJE5/action/replication_record"}},"created_at":"2026-06-25T00:18:16.635246+00:00","updated_at":"2026-06-25T00:18:16.635246+00:00"}