{"paper":{"title":"Block-wise Codeword Embedding for Reliable Multi-bit Text Watermarking","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Multi-bit LLM watermarking can reach 96.5 percent true positives at only 2 percent false positives by separating block-wise message estimation from window-shifting verification.","cross_cats":["cs.CL"],"primary_cat":"cs.CR","authors_text":"Dongsup Jin, HoEun Kim, Joeun Kim, Young-Sik Kim","submitted_at":"2026-05-01T02:14:38Z","abstract_excerpt":"Recent multi-bit watermarking methods for large language models (LLMs) prioritize capacity over reliability, often conflating decoding with detection. Our analysis reveals that existing ECC-based extractors suffer from catastrophic false positive rates (FPR), and applying rejection thresholds merely collapses detection sensitivity (TPR) to random guessing. To resolve this structural limitation, we propose BREW (Block-wise Reliable Embedding for Watermarking), a framework shifting the paradigm to designated verification. BREW employs a two-stage mechanism: (i) blind message estimation via indep"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"BREW achieves a TPR of 0.965 with an FPR of 0.02 under 10% synonym substitution, demonstrating that the high-FPR issue is not an inherent trade-off of multi-bit watermarking, but a solvable structural flaw of prior decoding-centric designs.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The two-stage mechanism of blind message estimation via independent block voting followed by window-shifting verification will rigorously validate the payload against local edits without introducing new failure modes or depending on unstated properties of the underlying LLM distribution.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"BREW achieves TPR of 0.965 and FPR of 0.02 under 10% synonym substitution by shifting from ECC decoding to designated verification with block voting and local validation.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Multi-bit LLM watermarking can reach 96.5 percent true positives at only 2 percent false positives by separating block-wise message estimation from window-shifting verification.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"b93ed408c18f70d046c26235e83451050cc8471d734ee87a55802ce99ea28909"},"source":{"id":"2605.00348","kind":"arxiv","version":2},"verdict":{"id":"a3784704-10ea-49a3-811c-face440ae33c","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-09T19:47:21.320421Z","strongest_claim":"BREW achieves a TPR of 0.965 with an FPR of 0.02 under 10% synonym substitution, demonstrating that the high-FPR issue is not an inherent trade-off of multi-bit watermarking, but a solvable structural flaw of prior decoding-centric designs.","one_line_summary":"BREW achieves TPR of 0.965 and FPR of 0.02 under 10% synonym substitution by shifting from ECC decoding to designated verification with block voting and local validation.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The two-stage mechanism of blind message estimation via independent block voting followed by window-shifting verification will rigorously validate the payload against local edits without introducing new failure modes or depending on unstated properties of the underlying LLM distribution.","pith_extraction_headline":"Multi-bit LLM watermarking can reach 96.5 percent true positives at only 2 percent false positives by separating block-wise message estimation from window-shifting verification."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.00348/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T20:34:05.716902Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T18:15:09.473497Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"ea0137a75a35335d6421602ebe27dce79a09b89064afabd2a75d2f30a6a6258a"},"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"}