{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:RG3CQAZEX2M5A25MLLGNXFWWFJ","short_pith_number":"pith:RG3CQAZE","canonical_record":{"source":{"id":"2411.19563","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-11-29T09:18:32Z","cross_cats_sorted":[],"title_canon_sha256":"9ed3c86415827315321116dcb16f08f707d1c02560c83aa873e9a42ad95b538c","abstract_canon_sha256":"0128c41f2f942a6a6ac957063d8bd0b34e6d9b7cedd521031d9dfa649a36733a"},"schema_version":"1.0"},"canonical_sha256":"89b6280324be99d06bac5accdb96d62a6dd69c04b827c8349fee0e4730a59f4c","source":{"kind":"arxiv","id":"2411.19563","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2411.19563","created_at":"2026-07-05T11:22:34Z"},{"alias_kind":"arxiv_version","alias_value":"2411.19563v2","created_at":"2026-07-05T11:22:34Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.19563","created_at":"2026-07-05T11:22:34Z"},{"alias_kind":"pith_short_12","alias_value":"RG3CQAZEX2M5","created_at":"2026-07-05T11:22:34Z"},{"alias_kind":"pith_short_16","alias_value":"RG3CQAZEX2M5A25M","created_at":"2026-07-05T11:22:34Z"},{"alias_kind":"pith_short_8","alias_value":"RG3CQAZE","created_at":"2026-07-05T11:22:34Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:RG3CQAZEX2M5A25MLLGNXFWWFJ","target":"record","payload":{"canonical_record":{"source":{"id":"2411.19563","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-11-29T09:18:32Z","cross_cats_sorted":[],"title_canon_sha256":"9ed3c86415827315321116dcb16f08f707d1c02560c83aa873e9a42ad95b538c","abstract_canon_sha256":"0128c41f2f942a6a6ac957063d8bd0b34e6d9b7cedd521031d9dfa649a36733a"},"schema_version":"1.0"},"canonical_sha256":"89b6280324be99d06bac5accdb96d62a6dd69c04b827c8349fee0e4730a59f4c","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:22:34.277825Z","signature_b64":"xBAop7qjQ93ORr5ZHEhbbo/2RsVEM+XkElb+SXXG0GEjxAdi9jGrI0jegb7Mi5Rvx3WxTeobHPOk34rynMctCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"89b6280324be99d06bac5accdb96d62a6dd69c04b827c8349fee0e4730a59f4c","last_reissued_at":"2026-07-05T11:22:34.277268Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:22:34.277268Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2411.19563","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:22:34Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"7vj9dFQSG0ETQXYeHfbBQealmgpb4iSs1I6l6/kFRfGZmwh7YJxNcxG1K1vo4zsxBBJoRrV/aYz/l+pE0/MYCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-14T17:54:17.984309Z"},"content_sha256":"e13b3ba410e969fd62b82ac3e6923a52f0bf716759fb50156d2176a4f6c81d9a","schema_version":"1.0","event_id":"sha256:e13b3ba410e969fd62b82ac3e6923a52f0bf716759fb50156d2176a4f6c81d9a"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:RG3CQAZEX2M5A25MLLGNXFWWFJ","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Ensemble Watermarks for Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Georg Niess, Roman Kern","submitted_at":"2024-11-29T09:18:32Z","abstract_excerpt":"As large language models (LLMs) reach human-like fluency, reliably distinguishing AI-generated text from human authorship becomes increasingly difficult. While watermarks already exist for LLMs, they often lack flexibility and struggle with attacks such as paraphrasing. To address these issues, we propose a multi-feature method for generating watermarks that combines multiple distinct watermark features into an ensemble watermark. Concretely, we combine acrostica and sensorimotor norms with the established red-green watermark to achieve a 98% detection rate. After a paraphrasing attack, the pe"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.19563","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/2411.19563/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:22:34Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"lRu+qNTyX+TQRcf4xq1BwDaf4sbHFCUMhWkxjvpvQXA7lnlAMBE09lf4J0o2G755R+0jpF90aPBNipjAcWYBBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-14T17:54:17.985101Z"},"content_sha256":"236e12ccad06aa8d160b8d982210489faea0926e0ab07e8eeda36a3bc02fa648","schema_version":"1.0","event_id":"sha256:236e12ccad06aa8d160b8d982210489faea0926e0ab07e8eeda36a3bc02fa648"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/RG3CQAZEX2M5A25MLLGNXFWWFJ/bundle.json","state_url":"https://pith.science/pith/RG3CQAZEX2M5A25MLLGNXFWWFJ/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/RG3CQAZEX2M5A25MLLGNXFWWFJ/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-14T17:54:18Z","links":{"resolver":"https://pith.science/pith/RG3CQAZEX2M5A25MLLGNXFWWFJ","bundle":"https://pith.science/pith/RG3CQAZEX2M5A25MLLGNXFWWFJ/bundle.json","state":"https://pith.science/pith/RG3CQAZEX2M5A25MLLGNXFWWFJ/state.json","well_known_bundle":"https://pith.science/.well-known/pith/RG3CQAZEX2M5A25MLLGNXFWWFJ/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:RG3CQAZEX2M5A25MLLGNXFWWFJ","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"0128c41f2f942a6a6ac957063d8bd0b34e6d9b7cedd521031d9dfa649a36733a","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-11-29T09:18:32Z","title_canon_sha256":"9ed3c86415827315321116dcb16f08f707d1c02560c83aa873e9a42ad95b538c"},"schema_version":"1.0","source":{"id":"2411.19563","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2411.19563","created_at":"2026-07-05T11:22:34Z"},{"alias_kind":"arxiv_version","alias_value":"2411.19563v2","created_at":"2026-07-05T11:22:34Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.19563","created_at":"2026-07-05T11:22:34Z"},{"alias_kind":"pith_short_12","alias_value":"RG3CQAZEX2M5","created_at":"2026-07-05T11:22:34Z"},{"alias_kind":"pith_short_16","alias_value":"RG3CQAZEX2M5A25M","created_at":"2026-07-05T11:22:34Z"},{"alias_kind":"pith_short_8","alias_value":"RG3CQAZE","created_at":"2026-07-05T11:22:34Z"}],"graph_snapshots":[{"event_id":"sha256:236e12ccad06aa8d160b8d982210489faea0926e0ab07e8eeda36a3bc02fa648","target":"graph","created_at":"2026-07-05T11:22:34Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2411.19563/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"As large language models (LLMs) reach human-like fluency, reliably distinguishing AI-generated text from human authorship becomes increasingly difficult. While watermarks already exist for LLMs, they often lack flexibility and struggle with attacks such as paraphrasing. To address these issues, we propose a multi-feature method for generating watermarks that combines multiple distinct watermark features into an ensemble watermark. Concretely, we combine acrostica and sensorimotor norms with the established red-green watermark to achieve a 98% detection rate. After a paraphrasing attack, the pe","authors_text":"Georg Niess, Roman Kern","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-11-29T09:18:32Z","title":"Ensemble Watermarks for Large Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.19563","kind":"arxiv","version":2},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:e13b3ba410e969fd62b82ac3e6923a52f0bf716759fb50156d2176a4f6c81d9a","target":"record","created_at":"2026-07-05T11:22:34Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"0128c41f2f942a6a6ac957063d8bd0b34e6d9b7cedd521031d9dfa649a36733a","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-11-29T09:18:32Z","title_canon_sha256":"9ed3c86415827315321116dcb16f08f707d1c02560c83aa873e9a42ad95b538c"},"schema_version":"1.0","source":{"id":"2411.19563","kind":"arxiv","version":2}},"canonical_sha256":"89b6280324be99d06bac5accdb96d62a6dd69c04b827c8349fee0e4730a59f4c","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"89b6280324be99d06bac5accdb96d62a6dd69c04b827c8349fee0e4730a59f4c","first_computed_at":"2026-07-05T11:22:34.277268Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:22:34.277268Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"xBAop7qjQ93ORr5ZHEhbbo/2RsVEM+XkElb+SXXG0GEjxAdi9jGrI0jegb7Mi5Rvx3WxTeobHPOk34rynMctCg==","signature_status":"signed_v1","signed_at":"2026-07-05T11:22:34.277825Z","signed_message":"canonical_sha256_bytes"},"source_id":"2411.19563","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:e13b3ba410e969fd62b82ac3e6923a52f0bf716759fb50156d2176a4f6c81d9a","sha256:236e12ccad06aa8d160b8d982210489faea0926e0ab07e8eeda36a3bc02fa648"],"state_sha256":"0df12623eb5f002c254d3e91630114f089f2d2556dcf226760394b349f562123"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"BwO+YGOj37nwSC+rCfBHZbGb/npF5gVfYV/HyBnLygS+BSe31mq6L10mCbp/FW+LVQKQMG3SfIBunpTFxsWZAg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-14T17:54:18.007633Z","bundle_sha256":"54bc91fb799b5d8bc2f5e429c52b7d8b24e683cb0a48df9b413ca8eea9afc3b5"}}