{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:V6GAP655GLJSY4UWCHDA44TOLS","short_pith_number":"pith:V6GAP655","schema_version":"1.0","canonical_sha256":"af8c07fbbd32d32c729611c60e726e5ca81b96e7e70d151b6160c244ee9def68","source":{"kind":"arxiv","id":"2403.14719","version":1},"attestation_state":"computed","paper":{"title":"Bypassing LLM Watermarks with Color-Aware Substitutions","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","cs.LG"],"primary_cat":"cs.CR","authors_text":"Qilong Wu, Varun Chandrasekaran","submitted_at":"2024-03-19T17:54:39Z","abstract_excerpt":"Watermarking approaches are proposed to identify if text being circulated is human or large language model (LLM) generated. The state-of-the-art watermarking strategy of Kirchenbauer et al. (2023a) biases the LLM to generate specific (``green'') tokens. However, determining the robustness of this watermarking method is an open problem. Existing attack methods fail to evade detection for longer text segments. We overcome this limitation, and propose {\\em Self Color Testing-based Substitution (SCTS)}, the first ``color-aware'' attack. SCTS obtains color information by strategically prompting the"},"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":"2403.14719","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2024-03-19T17:54:39Z","cross_cats_sorted":["cs.CV","cs.LG"],"title_canon_sha256":"7884f21fb4158a51252adf5b72c07aa96bf24415310634c33250f22597bf0ef2","abstract_canon_sha256":"d204435d8f2b76e394e50aed0b77b1973ae3398b767911d1db21f796e4718510"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:59:25.129190Z","signature_b64":"0G8h88UI0tadjTr0iQJpNuf0+VbAYw4ChUHL26cMJORIHu3C8A+VD9LbOOHGRWUlQhd4NtNUoE5bTweh9M89AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"af8c07fbbd32d32c729611c60e726e5ca81b96e7e70d151b6160c244ee9def68","last_reissued_at":"2026-07-05T07:59:25.128708Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:59:25.128708Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Bypassing LLM Watermarks with Color-Aware Substitutions","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","cs.LG"],"primary_cat":"cs.CR","authors_text":"Qilong Wu, Varun Chandrasekaran","submitted_at":"2024-03-19T17:54:39Z","abstract_excerpt":"Watermarking approaches are proposed to identify if text being circulated is human or large language model (LLM) generated. The state-of-the-art watermarking strategy of Kirchenbauer et al. (2023a) biases the LLM to generate specific (``green'') tokens. However, determining the robustness of this watermarking method is an open problem. Existing attack methods fail to evade detection for longer text segments. We overcome this limitation, and propose {\\em Self Color Testing-based Substitution (SCTS)}, the first ``color-aware'' attack. SCTS obtains color information by strategically prompting the"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.14719","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/2403.14719/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":"2403.14719","created_at":"2026-07-05T07:59:25.128772+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.14719v1","created_at":"2026-07-05T07:59:25.128772+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.14719","created_at":"2026-07-05T07:59:25.128772+00:00"},{"alias_kind":"pith_short_12","alias_value":"V6GAP655GLJS","created_at":"2026-07-05T07:59:25.128772+00:00"},{"alias_kind":"pith_short_16","alias_value":"V6GAP655GLJSY4UW","created_at":"2026-07-05T07:59:25.128772+00:00"},{"alias_kind":"pith_short_8","alias_value":"V6GAP655","created_at":"2026-07-05T07:59:25.128772+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2411.11478","citing_title":"SoK: On the Role and Future of AIGC Watermarking in the Era of Gen-AI","ref_index":131,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/V6GAP655GLJSY4UWCHDA44TOLS","json":"https://pith.science/pith/V6GAP655GLJSY4UWCHDA44TOLS.json","graph_json":"https://pith.science/api/pith-number/V6GAP655GLJSY4UWCHDA44TOLS/graph.json","events_json":"https://pith.science/api/pith-number/V6GAP655GLJSY4UWCHDA44TOLS/events.json","paper":"https://pith.science/paper/V6GAP655"},"agent_actions":{"view_html":"https://pith.science/pith/V6GAP655GLJSY4UWCHDA44TOLS","download_json":"https://pith.science/pith/V6GAP655GLJSY4UWCHDA44TOLS.json","view_paper":"https://pith.science/paper/V6GAP655","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.14719&json=true","fetch_graph":"https://pith.science/api/pith-number/V6GAP655GLJSY4UWCHDA44TOLS/graph.json","fetch_events":"https://pith.science/api/pith-number/V6GAP655GLJSY4UWCHDA44TOLS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/V6GAP655GLJSY4UWCHDA44TOLS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/V6GAP655GLJSY4UWCHDA44TOLS/action/storage_attestation","attest_author":"https://pith.science/pith/V6GAP655GLJSY4UWCHDA44TOLS/action/author_attestation","sign_citation":"https://pith.science/pith/V6GAP655GLJSY4UWCHDA44TOLS/action/citation_signature","submit_replication":"https://pith.science/pith/V6GAP655GLJSY4UWCHDA44TOLS/action/replication_record"}},"created_at":"2026-07-05T07:59:25.128772+00:00","updated_at":"2026-07-05T07:59:25.128772+00:00"}