{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:N46UMRFTT55WGREQMRDFIUEWG4","short_pith_number":"pith:N46UMRFT","schema_version":"1.0","canonical_sha256":"6f3d4644b39f7b6344906446545096371a2eb4ab18b6914c455e4acc7341f8bc","source":{"kind":"arxiv","id":"2410.12318","version":2},"attestation_state":"computed","paper":{"title":"UTF:Undertrained Tokens as Fingerprints A Novel Approach to LLM Identification","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Jiacheng Cai, Jiahao Yu, Yangguang Shao, Yuhang Wu","submitted_at":"2024-10-16T07:36:57Z","abstract_excerpt":"Fingerprinting large language models (LLMs) is essential for verifying model ownership, ensuring authenticity, and preventing misuse. Traditional fingerprinting methods often require significant computational overhead or white-box verification access. In this paper, we introduce UTF, a novel and efficient approach to fingerprinting LLMs by leveraging under-trained tokens. Under-trained tokens are tokens that the model has not fully learned during its training phase. By utilizing these tokens, we perform supervised fine-tuning to embed specific input-output pairs into the model. This process al"},"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":"2410.12318","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2024-10-16T07:36:57Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"43c85a21c5dfdbd67f8099a8235a4b4e6e4275051a476821cf91bfeb10b6a2ee","abstract_canon_sha256":"1462ea968980b7d98198bd5d6c3e4875f580183ba891f8e15a3dd0e547cd4654"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:35:51.601790Z","signature_b64":"nIa2uz+9M12RO+w/L3TYZF7Y01ldi3MhoE6pxvH5NCzZ4h+/ypJoL9ahdwC6gJlln63DAf/NxIoxrls+e7EiBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6f3d4644b39f7b6344906446545096371a2eb4ab18b6914c455e4acc7341f8bc","last_reissued_at":"2026-07-05T11:35:51.601366Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:35:51.601366Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"UTF:Undertrained Tokens as Fingerprints A Novel Approach to LLM Identification","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Jiacheng Cai, Jiahao Yu, Yangguang Shao, Yuhang Wu","submitted_at":"2024-10-16T07:36:57Z","abstract_excerpt":"Fingerprinting large language models (LLMs) is essential for verifying model ownership, ensuring authenticity, and preventing misuse. Traditional fingerprinting methods often require significant computational overhead or white-box verification access. In this paper, we introduce UTF, a novel and efficient approach to fingerprinting LLMs by leveraging under-trained tokens. Under-trained tokens are tokens that the model has not fully learned during its training phase. By utilizing these tokens, we perform supervised fine-tuning to embed specific input-output pairs into the model. This process al"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.12318","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/2410.12318/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":"2410.12318","created_at":"2026-07-05T11:35:51.601428+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.12318v2","created_at":"2026-07-05T11:35:51.601428+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.12318","created_at":"2026-07-05T11:35:51.601428+00:00"},{"alias_kind":"pith_short_12","alias_value":"N46UMRFTT55W","created_at":"2026-07-05T11:35:51.601428+00:00"},{"alias_kind":"pith_short_16","alias_value":"N46UMRFTT55WGREQ","created_at":"2026-07-05T11:35:51.601428+00:00"},{"alias_kind":"pith_short_8","alias_value":"N46UMRFT","created_at":"2026-07-05T11:35:51.601428+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2508.11548","citing_title":"Copyright Protection for Large Language Models: A Survey of Methods, Challenges, and Trends","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2510.18333","citing_title":"Position: LLM Watermarking Should Align Stakeholders' Incentives for Practical Adoption","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2604.05502","citing_title":"AttnDiff: Attention-based Differential Fingerprinting for Large Language Models","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/N46UMRFTT55WGREQMRDFIUEWG4","json":"https://pith.science/pith/N46UMRFTT55WGREQMRDFIUEWG4.json","graph_json":"https://pith.science/api/pith-number/N46UMRFTT55WGREQMRDFIUEWG4/graph.json","events_json":"https://pith.science/api/pith-number/N46UMRFTT55WGREQMRDFIUEWG4/events.json","paper":"https://pith.science/paper/N46UMRFT"},"agent_actions":{"view_html":"https://pith.science/pith/N46UMRFTT55WGREQMRDFIUEWG4","download_json":"https://pith.science/pith/N46UMRFTT55WGREQMRDFIUEWG4.json","view_paper":"https://pith.science/paper/N46UMRFT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.12318&json=true","fetch_graph":"https://pith.science/api/pith-number/N46UMRFTT55WGREQMRDFIUEWG4/graph.json","fetch_events":"https://pith.science/api/pith-number/N46UMRFTT55WGREQMRDFIUEWG4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/N46UMRFTT55WGREQMRDFIUEWG4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/N46UMRFTT55WGREQMRDFIUEWG4/action/storage_attestation","attest_author":"https://pith.science/pith/N46UMRFTT55WGREQMRDFIUEWG4/action/author_attestation","sign_citation":"https://pith.science/pith/N46UMRFTT55WGREQMRDFIUEWG4/action/citation_signature","submit_replication":"https://pith.science/pith/N46UMRFTT55WGREQMRDFIUEWG4/action/replication_record"}},"created_at":"2026-07-05T11:35:51.601428+00:00","updated_at":"2026-07-05T11:35:51.601428+00:00"}