{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:RR4OOL7WJ3GAYIX4KSMA2OD42N","short_pith_number":"pith:RR4OOL7W","schema_version":"1.0","canonical_sha256":"8c78e72ff64ecc0c22fc54980d387cd37e1bf3f2c623bd21564a01acd7e7844a","source":{"kind":"arxiv","id":"2402.12948","version":3},"attestation_state":"computed","paper":{"title":"GumbelSoft: Diversified Language Model Watermarking via the GumbelMax-trick","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jiangjie Chen, Jiayi Fu, Ruihan Yang, Xuandong Zhao, Yanghua Xiao, Yuansen Zhang","submitted_at":"2024-02-20T12:05:47Z","abstract_excerpt":"Large language models (LLMs) excellently generate human-like text, but also raise concerns about misuse in fake news and academic dishonesty. Decoding-based watermark, particularly the GumbelMax-trick-based watermark(GM watermark), is a standout solution for safeguarding machine-generated texts due to its notable detectability. However, GM watermark encounters a major challenge with generation diversity, always yielding identical outputs for the same prompt, negatively impacting generation diversity and user experience. To overcome this limitation, we propose a new type of GM watermark, the Lo"},"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":"2402.12948","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-02-20T12:05:47Z","cross_cats_sorted":[],"title_canon_sha256":"af28d8039d644cf108aa4e3c4b4bf414a3a3b46670582b87f91d1cfde94abc6d","abstract_canon_sha256":"7533b42f646733de490dad74687716efdef2120626d28c68574e2b128f308181"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:24:02.375298Z","signature_b64":"NCZpiZP5wfrETDs+MItGcDBEin9v17WVyJztc823XyTjNMwJ5kL8D6o6XR1zerf9yRxh/6Bp2ERuOR1TYF75BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8c78e72ff64ecc0c22fc54980d387cd37e1bf3f2c623bd21564a01acd7e7844a","last_reissued_at":"2026-07-05T08:24:02.374811Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:24:02.374811Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"GumbelSoft: Diversified Language Model Watermarking via the GumbelMax-trick","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jiangjie Chen, Jiayi Fu, Ruihan Yang, Xuandong Zhao, Yanghua Xiao, Yuansen Zhang","submitted_at":"2024-02-20T12:05:47Z","abstract_excerpt":"Large language models (LLMs) excellently generate human-like text, but also raise concerns about misuse in fake news and academic dishonesty. Decoding-based watermark, particularly the GumbelMax-trick-based watermark(GM watermark), is a standout solution for safeguarding machine-generated texts due to its notable detectability. However, GM watermark encounters a major challenge with generation diversity, always yielding identical outputs for the same prompt, negatively impacting generation diversity and user experience. To overcome this limitation, we propose a new type of GM watermark, the Lo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.12948","kind":"arxiv","version":3},"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/2402.12948/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":"2402.12948","created_at":"2026-07-05T08:24:02.374873+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.12948v3","created_at":"2026-07-05T08:24:02.374873+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.12948","created_at":"2026-07-05T08:24:02.374873+00:00"},{"alias_kind":"pith_short_12","alias_value":"RR4OOL7WJ3GA","created_at":"2026-07-05T08:24:02.374873+00:00"},{"alias_kind":"pith_short_16","alias_value":"RR4OOL7WJ3GAYIX4","created_at":"2026-07-05T08:24:02.374873+00:00"},{"alias_kind":"pith_short_8","alias_value":"RR4OOL7W","created_at":"2026-07-05T08:24:02.374873+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.05353","citing_title":"Selective Disclosure Watermarking for Large Language Models","ref_index":14,"is_internal_anchor":true},{"citing_arxiv_id":"2510.18333","citing_title":"Position: LLM Watermarking Should Align Stakeholders' Incentives for Practical Adoption","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08964","citing_title":"Trustworthy AI: Ensuring Reliability and Accountability from Models to Agents","ref_index":53,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RR4OOL7WJ3GAYIX4KSMA2OD42N","json":"https://pith.science/pith/RR4OOL7WJ3GAYIX4KSMA2OD42N.json","graph_json":"https://pith.science/api/pith-number/RR4OOL7WJ3GAYIX4KSMA2OD42N/graph.json","events_json":"https://pith.science/api/pith-number/RR4OOL7WJ3GAYIX4KSMA2OD42N/events.json","paper":"https://pith.science/paper/RR4OOL7W"},"agent_actions":{"view_html":"https://pith.science/pith/RR4OOL7WJ3GAYIX4KSMA2OD42N","download_json":"https://pith.science/pith/RR4OOL7WJ3GAYIX4KSMA2OD42N.json","view_paper":"https://pith.science/paper/RR4OOL7W","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.12948&json=true","fetch_graph":"https://pith.science/api/pith-number/RR4OOL7WJ3GAYIX4KSMA2OD42N/graph.json","fetch_events":"https://pith.science/api/pith-number/RR4OOL7WJ3GAYIX4KSMA2OD42N/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RR4OOL7WJ3GAYIX4KSMA2OD42N/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RR4OOL7WJ3GAYIX4KSMA2OD42N/action/storage_attestation","attest_author":"https://pith.science/pith/RR4OOL7WJ3GAYIX4KSMA2OD42N/action/author_attestation","sign_citation":"https://pith.science/pith/RR4OOL7WJ3GAYIX4KSMA2OD42N/action/citation_signature","submit_replication":"https://pith.science/pith/RR4OOL7WJ3GAYIX4KSMA2OD42N/action/replication_record"}},"created_at":"2026-07-05T08:24:02.374873+00:00","updated_at":"2026-07-05T08:24:02.374873+00:00"}