{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:VYZQZ7TG3S2MIDSQFCQSPW3AI5","short_pith_number":"pith:VYZQZ7TG","schema_version":"1.0","canonical_sha256":"ae330cfe66dcb4c40e5028a127db6047409dfd946ec8a3e63c5d3e32a5a71db3","source":{"kind":"arxiv","id":"2504.16046","version":2},"attestation_state":"computed","paper":{"title":"Certified Mitigation of Worst-Case LLM Copyright Infringement","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Benjamin Van Durme, Daniel Khashabi, Jiacan Yu, Jingyu Zhang, Marc Marone","submitted_at":"2025-04-22T17:16:53Z","abstract_excerpt":"The exposure of large language models (LLMs) to copyrighted material during pre-training raises concerns about unintentional copyright infringement post deployment. This has driven the development of \"copyright takedown\" methods, post-training approaches aimed at preventing models from generating content substantially similar to copyrighted ones. While current mitigation approaches are somewhat effective for average-case risks, we demonstrate that they overlook worst-case copyright risks exhibits by the existence of long, verbatim quotes from copyrighted sources. We propose BloomScrub, a remar"},"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":"2504.16046","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-04-22T17:16:53Z","cross_cats_sorted":[],"title_canon_sha256":"79ed01da26da4b3487ab5939d297658aceee1de5acde96d5e22ec1e0c4a8fd05","abstract_canon_sha256":"88aab64f9c51ce99b90eff0a267fb85667eb1f7c68b1dcb19d57088636113aa1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:53:00.226797Z","signature_b64":"glKQVqQXTtta8KtNXlmgvOKJ+QRFp/9VSp8qNFnytdtHmTvWUTD4Rw2Z4voY9dQCCLHJAa+IUkSKInaxT3qQCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ae330cfe66dcb4c40e5028a127db6047409dfd946ec8a3e63c5d3e32a5a71db3","last_reissued_at":"2026-07-05T10:53:00.226387Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:53:00.226387Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Certified Mitigation of Worst-Case LLM Copyright Infringement","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Benjamin Van Durme, Daniel Khashabi, Jiacan Yu, Jingyu Zhang, Marc Marone","submitted_at":"2025-04-22T17:16:53Z","abstract_excerpt":"The exposure of large language models (LLMs) to copyrighted material during pre-training raises concerns about unintentional copyright infringement post deployment. This has driven the development of \"copyright takedown\" methods, post-training approaches aimed at preventing models from generating content substantially similar to copyrighted ones. While current mitigation approaches are somewhat effective for average-case risks, we demonstrate that they overlook worst-case copyright risks exhibits by the existence of long, verbatim quotes from copyrighted sources. We propose BloomScrub, a remar"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.16046","kind":"arxiv","version":2},"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/2504.16046/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":"2504.16046","created_at":"2026-07-05T10:53:00.226444+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.16046v2","created_at":"2026-07-05T10:53:00.226444+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.16046","created_at":"2026-07-05T10:53:00.226444+00:00"},{"alias_kind":"pith_short_12","alias_value":"VYZQZ7TG3S2M","created_at":"2026-07-05T10:53:00.226444+00:00"},{"alias_kind":"pith_short_16","alias_value":"VYZQZ7TG3S2MIDSQ","created_at":"2026-07-05T10:53:00.226444+00:00"},{"alias_kind":"pith_short_8","alias_value":"VYZQZ7TG","created_at":"2026-07-05T10:53:00.226444+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.03547","citing_title":"Erase Persona, Forget Lore: Benchmarking Multimodal Copyright Unlearning in Large Vision Language Models","ref_index":38,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VYZQZ7TG3S2MIDSQFCQSPW3AI5","json":"https://pith.science/pith/VYZQZ7TG3S2MIDSQFCQSPW3AI5.json","graph_json":"https://pith.science/api/pith-number/VYZQZ7TG3S2MIDSQFCQSPW3AI5/graph.json","events_json":"https://pith.science/api/pith-number/VYZQZ7TG3S2MIDSQFCQSPW3AI5/events.json","paper":"https://pith.science/paper/VYZQZ7TG"},"agent_actions":{"view_html":"https://pith.science/pith/VYZQZ7TG3S2MIDSQFCQSPW3AI5","download_json":"https://pith.science/pith/VYZQZ7TG3S2MIDSQFCQSPW3AI5.json","view_paper":"https://pith.science/paper/VYZQZ7TG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.16046&json=true","fetch_graph":"https://pith.science/api/pith-number/VYZQZ7TG3S2MIDSQFCQSPW3AI5/graph.json","fetch_events":"https://pith.science/api/pith-number/VYZQZ7TG3S2MIDSQFCQSPW3AI5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VYZQZ7TG3S2MIDSQFCQSPW3AI5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VYZQZ7TG3S2MIDSQFCQSPW3AI5/action/storage_attestation","attest_author":"https://pith.science/pith/VYZQZ7TG3S2MIDSQFCQSPW3AI5/action/author_attestation","sign_citation":"https://pith.science/pith/VYZQZ7TG3S2MIDSQFCQSPW3AI5/action/citation_signature","submit_replication":"https://pith.science/pith/VYZQZ7TG3S2MIDSQFCQSPW3AI5/action/replication_record"}},"created_at":"2026-07-05T10:53:00.226444+00:00","updated_at":"2026-07-05T10:53:00.226444+00:00"}