{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:MPKMKJTZKB6TL4XRW67OC2NLVH","short_pith_number":"pith:MPKMKJTZ","schema_version":"1.0","canonical_sha256":"63d4c52679507d35f2f1b7bee169aba9c507e820c7be833e47e8f21f87bc013d","source":{"kind":"arxiv","id":"2411.11203","version":2},"attestation_state":"computed","paper":{"title":"Debiasing Watermarks for Large Language Models via Maximal Coupling","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.CR","cs.LG","stat.ME"],"primary_cat":"stat.ML","authors_text":"Ruixun Zhang, Tanwi Mallick, Weijie J. Su, Xiang Li, Yangxinyu Xie","submitted_at":"2024-11-17T23:36:37Z","abstract_excerpt":"Watermarking language models is essential for distinguishing between human and machine-generated text and thus maintaining the integrity and trustworthiness of digital communication. We present a novel green/red list watermarking approach that partitions the token set into ``green'' and ``red'' lists, subtly increasing the generation probability for green tokens. To correct token distribution bias, our method employs maximal coupling, using a uniform coin flip to decide whether to apply bias correction, with the result embedded as a pseudorandom watermark signal. Theoretical analysis confirms "},"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":"2411.11203","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2024-11-17T23:36:37Z","cross_cats_sorted":["cs.CL","cs.CR","cs.LG","stat.ME"],"title_canon_sha256":"c26c4a54f21f51691166704df06bbe1529a0361b92a194c43869aad0316ebf5b","abstract_canon_sha256":"860a17b268184e17449ce21995488f99c0fed068b67c9a8e59605ac51c0e9e2c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:20:37.066545Z","signature_b64":"wpSkHVUT2xxtrLLdD73vhtGSqpC0Gzwrsfdko4OtXCA7Jcg/TI9HWdFjxPUMYeJhS6n3HIBxlLTmrocIIo4NDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"63d4c52679507d35f2f1b7bee169aba9c507e820c7be833e47e8f21f87bc013d","last_reissued_at":"2026-07-05T11:20:37.066043Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:20:37.066043Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Debiasing Watermarks for Large Language Models via Maximal Coupling","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.CR","cs.LG","stat.ME"],"primary_cat":"stat.ML","authors_text":"Ruixun Zhang, Tanwi Mallick, Weijie J. Su, Xiang Li, Yangxinyu Xie","submitted_at":"2024-11-17T23:36:37Z","abstract_excerpt":"Watermarking language models is essential for distinguishing between human and machine-generated text and thus maintaining the integrity and trustworthiness of digital communication. We present a novel green/red list watermarking approach that partitions the token set into ``green'' and ``red'' lists, subtly increasing the generation probability for green tokens. To correct token distribution bias, our method employs maximal coupling, using a uniform coin flip to decide whether to apply bias correction, with the result embedded as a pseudorandom watermark signal. Theoretical analysis confirms "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.11203","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/2411.11203/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":"2411.11203","created_at":"2026-07-05T11:20:37.066103+00:00"},{"alias_kind":"arxiv_version","alias_value":"2411.11203v2","created_at":"2026-07-05T11:20:37.066103+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.11203","created_at":"2026-07-05T11:20:37.066103+00:00"},{"alias_kind":"pith_short_12","alias_value":"MPKMKJTZKB6T","created_at":"2026-07-05T11:20:37.066103+00:00"},{"alias_kind":"pith_short_16","alias_value":"MPKMKJTZKB6TL4XR","created_at":"2026-07-05T11:20:37.066103+00:00"},{"alias_kind":"pith_short_8","alias_value":"MPKMKJTZ","created_at":"2026-07-05T11:20:37.066103+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.08964","citing_title":"Trustworthy AI: Ensuring Reliability and Accountability from Models to Agents","ref_index":163,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MPKMKJTZKB6TL4XRW67OC2NLVH","json":"https://pith.science/pith/MPKMKJTZKB6TL4XRW67OC2NLVH.json","graph_json":"https://pith.science/api/pith-number/MPKMKJTZKB6TL4XRW67OC2NLVH/graph.json","events_json":"https://pith.science/api/pith-number/MPKMKJTZKB6TL4XRW67OC2NLVH/events.json","paper":"https://pith.science/paper/MPKMKJTZ"},"agent_actions":{"view_html":"https://pith.science/pith/MPKMKJTZKB6TL4XRW67OC2NLVH","download_json":"https://pith.science/pith/MPKMKJTZKB6TL4XRW67OC2NLVH.json","view_paper":"https://pith.science/paper/MPKMKJTZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2411.11203&json=true","fetch_graph":"https://pith.science/api/pith-number/MPKMKJTZKB6TL4XRW67OC2NLVH/graph.json","fetch_events":"https://pith.science/api/pith-number/MPKMKJTZKB6TL4XRW67OC2NLVH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MPKMKJTZKB6TL4XRW67OC2NLVH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MPKMKJTZKB6TL4XRW67OC2NLVH/action/storage_attestation","attest_author":"https://pith.science/pith/MPKMKJTZKB6TL4XRW67OC2NLVH/action/author_attestation","sign_citation":"https://pith.science/pith/MPKMKJTZKB6TL4XRW67OC2NLVH/action/citation_signature","submit_replication":"https://pith.science/pith/MPKMKJTZKB6TL4XRW67OC2NLVH/action/replication_record"}},"created_at":"2026-07-05T11:20:37.066103+00:00","updated_at":"2026-07-05T11:20:37.066103+00:00"}