{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:3UQLDSBGAMXIEW4PNG5P4CUB7J","short_pith_number":"pith:3UQLDSBG","schema_version":"1.0","canonical_sha256":"dd20b1c826032e825b8f69bafe0a81fa7a32246dda7d07003d13ca09dbf134ba","source":{"kind":"arxiv","id":"2504.12108","version":1},"attestation_state":"computed","paper":{"title":"Entropy-Guided Watermarking for LLMs: A Test-Time Framework for Robust and Traceable Text Generation","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Dacheng Tao, Liang Ding, Shizhan Cai","submitted_at":"2025-04-16T14:16:38Z","abstract_excerpt":"The rapid development of Large Language Models (LLMs) has intensified concerns about content traceability and potential misuse. Existing watermarking schemes for sampled text often face trade-offs between maintaining text quality and ensuring robust detection against various attacks. To address these issues, we propose a novel watermarking scheme that improves both detectability and text quality by introducing a cumulative watermark entropy threshold. Our approach is compatible with and generalizes existing sampling functions, enhancing adaptability. Experimental results across multiple LLMs s"},"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":"2504.12108","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.CL","submitted_at":"2025-04-16T14:16:38Z","cross_cats_sorted":[],"title_canon_sha256":"9b1511a703a1dfd3025fd1c97e45114c532fa60c38014e7e345dad3c3ae4191f","abstract_canon_sha256":"5460e50c8a12ee3988c9bd9cc12b671002698a590250ce7718acfaed6d749ae3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:50:02.712074Z","signature_b64":"mg8pC0zstOwa6ZF1K9sifoLQGhX+r3PMk9Fguqrhr+nMP/eziReoWKstSTHJEb0HhVW9zwr/+TZOrJp8FoCrAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dd20b1c826032e825b8f69bafe0a81fa7a32246dda7d07003d13ca09dbf134ba","last_reissued_at":"2026-07-05T10:50:02.711609Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:50:02.711609Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Entropy-Guided Watermarking for LLMs: A Test-Time Framework for Robust and Traceable Text Generation","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Dacheng Tao, Liang Ding, Shizhan Cai","submitted_at":"2025-04-16T14:16:38Z","abstract_excerpt":"The rapid development of Large Language Models (LLMs) has intensified concerns about content traceability and potential misuse. Existing watermarking schemes for sampled text often face trade-offs between maintaining text quality and ensuring robust detection against various attacks. To address these issues, we propose a novel watermarking scheme that improves both detectability and text quality by introducing a cumulative watermark entropy threshold. Our approach is compatible with and generalizes existing sampling functions, enhancing adaptability. Experimental results across multiple LLMs s"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.12108","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/2504.12108/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":"2504.12108","created_at":"2026-07-05T10:50:02.711665+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.12108v1","created_at":"2026-07-05T10:50:02.711665+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.12108","created_at":"2026-07-05T10:50:02.711665+00:00"},{"alias_kind":"pith_short_12","alias_value":"3UQLDSBGAMXI","created_at":"2026-07-05T10:50:02.711665+00:00"},{"alias_kind":"pith_short_16","alias_value":"3UQLDSBGAMXIEW4P","created_at":"2026-07-05T10:50:02.711665+00:00"},{"alias_kind":"pith_short_8","alias_value":"3UQLDSBG","created_at":"2026-07-05T10:50:02.711665+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2608.06416","citing_title":"WorldMark: A Plug-and-Play World Knowledge Interface for Cross-Host Language Model Watermarking","ref_index":15,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3UQLDSBGAMXIEW4PNG5P4CUB7J","json":"https://pith.science/pith/3UQLDSBGAMXIEW4PNG5P4CUB7J.json","graph_json":"https://pith.science/api/pith-number/3UQLDSBGAMXIEW4PNG5P4CUB7J/graph.json","events_json":"https://pith.science/api/pith-number/3UQLDSBGAMXIEW4PNG5P4CUB7J/events.json","paper":"https://pith.science/paper/3UQLDSBG"},"agent_actions":{"view_html":"https://pith.science/pith/3UQLDSBGAMXIEW4PNG5P4CUB7J","download_json":"https://pith.science/pith/3UQLDSBGAMXIEW4PNG5P4CUB7J.json","view_paper":"https://pith.science/paper/3UQLDSBG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.12108&json=true","fetch_graph":"https://pith.science/api/pith-number/3UQLDSBGAMXIEW4PNG5P4CUB7J/graph.json","fetch_events":"https://pith.science/api/pith-number/3UQLDSBGAMXIEW4PNG5P4CUB7J/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3UQLDSBGAMXIEW4PNG5P4CUB7J/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3UQLDSBGAMXIEW4PNG5P4CUB7J/action/storage_attestation","attest_author":"https://pith.science/pith/3UQLDSBGAMXIEW4PNG5P4CUB7J/action/author_attestation","sign_citation":"https://pith.science/pith/3UQLDSBGAMXIEW4PNG5P4CUB7J/action/citation_signature","submit_replication":"https://pith.science/pith/3UQLDSBGAMXIEW4PNG5P4CUB7J/action/replication_record"}},"created_at":"2026-07-05T10:50:02.711665+00:00","updated_at":"2026-07-05T10:50:02.711665+00:00"}