{"as_of":"2026-08-10T05:11:00Z","caps":{"database_statements":6,"inbound":100,"outbound":100},"context_digest":"sha256:a0dc02197edf681c27a5f8e8344ccfaf549d2144c7d531e8a79040b2852cfd14","coverage":[{"denominator":76,"lane":"reference_resolution","note":"Typed states for the displayed outbound observations.","records_observed":76,"source":"paper_references, paper_reference_links","source_observed_at":"2026-08-04T19:46:51.604848Z","state":"measured"},{"denominator":77,"lane":"standing_notices","note":"One-hop event checks from named stored sources.","records_observed":77,"source":"scholarly_work_events, retraction_status_cache","source_observed_at":"2026-08-09T06:31:02.800959+00:00","state":"measured"},{"denominator":1,"lane":"inbound_itemization","note":"Pith citing papers itemized under the disclosed page cap.","records_observed":1,"source":"paper_references, paper_reference_links","source_observed_at":"2026-08-06T04:16:06.800914Z","state":"measured"},{"denominator":1,"lane":"external_citation_measurements","note":"A source-named dated measurement, never combined with another source.","records_observed":0,"source":"pith","source_observed_at":"2026-08-06T04:16:07.072616Z","state":"measured"}],"external_citation_measurements":[],"inbound":[{"citation":{"cited_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"cited_work":{"arxiv_id":"2509.09112","doi":null,"metadata_source":"pith","pith_arxiv_id":"2509.09112","snapshot_observed_at":"2026-08-06T04:16:07.072616Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","venue":"cs.CR","work_id":"783b6d51-37a0-40a4-8f4a-4be4dae10cb0","year":2025},"citing_paper":{"arxiv_id":"2508.03836","last_updated":"2025-08-05T18:34:00Z","snapshot_observed_at":"2026-08-09T15:06:49.621621Z","submitted_at":"2025-08-05T18:34:00Z","title":"DP-NCB: Privacy Preserving Fair Bandits","version":1},"reference_index":44,"source":"pdf_text","source_observed_at":"2026-08-06T04:16:06.800914Z"},"links":{"cited_paper":"/paper/2509.09112","citing_paper":"/paper/2508.03836"},"observation_digest":"sha256:338c1560c7e4f6b9167961a6c356c1aaa5cb0fb8513c2f89abe4bcdb3a2f7b08","observation_id":"2ab31816-21f1-4680-ae47-4be4ff3b9c28","resolution":{"observed_at":"2026-08-06T04:16:07.155446Z","resolver_source":"local_arxiv","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-09T06:31:02.800959+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-09T06:31:02.800959+00:00","source":"crossref"},{"observed_at":"2026-08-09T06:30:57.326959+00:00","source":"retraction_watch"}],"state":"measured"}}],"links":{"evidence":"/evidence","html":"/paper/2509.09112/citation-record","integrity":"/paper/2509.09112/integrity","json":"/paper/2509.09112/citation-record.json","paper":"/paper/2509.09112"},"outbound":[{"citation":{"cited_paper":{"arxiv_id":"2205.01068","last_updated":"2022-06-21T17:04:40Z","snapshot_observed_at":"2026-08-06T03:13:37.403059Z","submitted_at":"2022-05-02T17:49:50Z","title":"OPT: Open Pre-trained Transformer Language Models","version":4},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2205.01068","snapshot_observed_at":"2026-08-04T19:46:49.866067Z","title":"Opt: Open pre-trained transformer language models,","venue":null,"work_id":null,"year":2022},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":1,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:49.866067Z"},"links":{"cited_paper":"/paper/2205.01068","citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:8f4b674b38ae30e92841a7459bfb376e2e3a8a2a74e9d61522b9d8c970d9c7b3","observation_id":"af1c3dac-5a6b-47f0-b9da-6114394187e5","resolution":{"observed_at":"2026-08-04T19:46:49.866067Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2302.13971","last_updated":"2023-02-27T17:11:15Z","snapshot_observed_at":"2026-07-06T02:11:23.670680Z","submitted_at":"2023-02-27T17:11:15Z","title":"LLaMA: Open and Efficient Foundation Language Models","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2302.13971","snapshot_observed_at":"2026-08-04T19:46:49.916419Z","title":"Llama: Open and efficient foundation language models,","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":2,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:49.916419Z"},"links":{"cited_paper":"/paper/2302.13971","citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:be58faf57b7e1ca27a38d5a21db58622f9217272f7fbccc94b4420588776c934","observation_id":"f95d0745-38c1-45f7-bf9d-1747d668a072","resolution":{"observed_at":"2026-08-04T19:46:49.916419Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:49.971897Z","title":"Openai chatgpt blog,","venue":null,"work_id":null,"year":2022},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":3,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:49.971897Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:9d4973ec9d65bb07abbafaea50e13c5d3f685074b66cd4c9e8711b262795d2a5","observation_id":"9e718246-4dfd-400e-9ab1-817f5907d1cc","resolution":{"observed_at":"2026-08-04T19:46:49.971897Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:50.024816Z","title":"Deepseek terms of use,","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":4,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.024816Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:f712188fc3f6bd0232983bd62eeaccba33fdba1b1c1506e7ccca8e6bf338be41","observation_id":"aeba42fb-1191-40ac-8275-8b3062c3f54a","resolution":{"observed_at":"2026-08-04T19:46:50.024816Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:50.054977Z","title":"Gemini api additional terms of service,","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":5,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.054977Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:0ee95a6df9f6c160f35d9d8eb3c388db1920d3b2ec375355cb4d552a2af82c71","observation_id":"73bbabcb-5d4a-413d-9124-68b9f30a9271","resolution":{"observed_at":"2026-08-04T19:46:50.054977Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:50.106363Z","title":"Grammarly","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":6,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.106363Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:69e5b0c8d835672177f2d8273ab33fc9fd9f61102fe8be7017d3aa712a261c3e","observation_id":"0dc0cfca-559d-4ebe-b84c-c1bbeb95ca34","resolution":{"observed_at":"2026-08-04T19:46:50.106363Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:50.136232Z","title":"Chatpaper","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":7,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.136232Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:655c2256dec8e8cb90af675e5415a3cbbb4a8a749b08049c3baab2f8e7db3f9c","observation_id":"86c47f59-411a-45c1-b8f9-03162a0c3751","resolution":{"observed_at":"2026-08-04T19:46:50.136232Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:50.189600Z","title":"Transforming the future of music creation","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":8,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.189600Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:67be118758e07765f3bc60daa9863a6eba7617271664529131c3aab510ede8db","observation_id":"25a8acb9-e072-476f-a2ce-7293e0584fa2","resolution":{"observed_at":"2026-08-04T19:46:50.189600Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:50.240567Z","title":"Quickstart: Generate and verify an image’s watermark using imagen text-to-image (console)","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":9,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.240567Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:f118638e3f60d764a94e9cf88dfe7b219eb6b4c4e435b80d2c33901cbc88c463","observation_id":"23e1971f-30a2-44ef-896a-8681f59f4373","resolution":{"observed_at":"2026-08-04T19:46:50.240567Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:50.243147Z","title":"Self-supervised adversarial example detection by disentangled representation,","venue":null,"work_id":null,"year":2022},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":10,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.243147Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:45d7182daaf39cea90f6913953a6a1544e8eb2c6f2675128fc4ba377d27e954e","observation_id":"610cf1b5-f003-4000-9641-276426b4deb3","resolution":{"observed_at":"2026-08-04T19:46:50.243147Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:50.266271Z","title":"Evaluating membership inference through adversarial robustness,","venue":null,"work_id":null,"year":2022},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":11,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.266271Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:a91fb2c000abf910a75386467e472ddb12e1ae9def433ff1c204f1a5cacbc3bd","observation_id":"3b04871c-e940-44a5-8b3b-f7d8352b813c","resolution":{"observed_at":"2026-08-04T19:46:50.266271Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:50.319803Z","title":"Not all edges are equally robust: Evaluating the robustness of ranking-based federated learning,","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":12,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.319803Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:34b5a8236652d8352d2cac594195eb45d21f95b04f5060b7683d7a9380f63120","observation_id":"64306f4c-6f26-478c-a3af-4a189294a195","resolution":{"observed_at":"2026-08-04T19:46:50.319803Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:50.387501Z","title":"When better features mean greater risks: The performance- privacy trade-off in contrastive learning,","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":13,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.387501Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:7b32e7b7b4c04412742c26dc1b4ff52d86338fe1052941ed5a93f83e2a3327c1","observation_id":"ca756dd5-3883-4cd9-92c7-7c9bd68bbd06","resolution":{"observed_at":"2026-08-04T19:46:50.387501Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:50.427491Z","title":"Dynamic graph unlearning: a general and efficient post-processing method via gradi- ent transformation,","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":14,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.427491Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:836f5f357b82519f6cd9c4a165e0b37d0f878444e2e8f6b08e0c040ddd84aa67","observation_id":"1f4cd71c-4cb2-4b1b-a515-e188907bf284","resolution":{"observed_at":"2026-08-04T19:46:50.427491Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:50.485556Z","title":"Unraveling privacy risks of individual fairness in graph neural networks,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":15,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.485556Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:b64258fa9e6ffaf80fcf3be29a835db82855b64173648aea015808ab7c8c86fa","observation_id":"b946c5b4-04e9-4458-ba16-ee32c629e76b","resolution":{"observed_at":"2026-08-04T19:46:50.485556Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:50.531624Z","title":"On the risk of misinformation pollution with large language models,","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":16,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.531624Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:4b04d19c3013ae1799704492d3d188a0ad5c118124184a8eb1698e7079b1b36f","observation_id":"4d14f793-8ff2-45e7-b6d5-339e9e2d1425","resolution":{"observed_at":"2026-08-04T19:46:50.531624Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2305.06972","last_updated":"2023-12-22T18:01:44Z","snapshot_observed_at":"2026-07-06T15:26:06.998516Z","submitted_at":"2023-05-11T16:55:19Z","title":"Spear Phishing With Large Language Models","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2305.06972","snapshot_observed_at":"2026-08-04T19:46:50.589714Z","title":"Large language models can be used to effectively scale spear phishing campaigns,","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":17,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.589714Z"},"links":{"cited_paper":"/paper/2305.06972","citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:56b65d5353d52d94eee7984a6e2d39bf42ff39454b5c1cc02d5887e9b1940c73","observation_id":"cd411e46-a3fb-49d2-8a12-4bdd5d8ab87c","resolution":{"observed_at":"2026-08-04T19:46:50.589714Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:50.647691Z","title":"Chatgpt for good? on opportunities and challenges of large language models for education,","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":18,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.647691Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:aac263d0b65634a0aff1013839f588d6cce832f6d2b65e5de3d89c0a30fa3cf9","observation_id":"dd80004e-74c3-433e-bef1-7e8fcf9fe996","resolution":{"observed_at":"2026-08-04T19:46:50.647691Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:50.687781Z","title":"A watermark for large language models,","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":19,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.687781Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:0435d17a2307c47c03be9077bc21284604cc39d92c96a7b5c2be98c15853baaf","observation_id":"6ef2c037-1a1a-42fe-9191-1cfbe991e1b9","resolution":{"observed_at":"2026-08-04T19:46:50.687781Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:50.734763Z","title":"Provable robust watermarking for AI-generated text,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":20,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.734763Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:f0eaa4f331b47c4768cd709bf3ac5932181c289d39276aae3a05c120e7696699","observation_id":"1f471d1d-fafd-4e8d-98d9-9a1f59772796","resolution":{"observed_at":"2026-08-04T19:46:50.734763Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:50.839856Z","title":"Unbiased watermark for large language models,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":21,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:50.839856Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:6564b7f374cd45b9c08d6707ced49deb219c1659b65ed61e9c90cd9f75ab81f0","observation_id":"7fba4bf8-80ec-4bab-b539-049ecb010e5d","resolution":{"observed_at":"2026-08-04T19:46:50.839856Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.012068Z","title":"A resilient and accessible distribution-preserving watermark for large language models,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":22,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.012068Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:25b32d1c06b70d24340e2ec4be7c54b2a2559b12869b5b23e516b1e2be95d327","observation_id":"681c2d0a-2566-4b9d-b959-ba06eb95b302","resolution":{"observed_at":"2026-08-04T19:46:51.012068Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.120204Z","title":"Scalable watermarking for identifying large language model outputs,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":23,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.120204Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:931fab6b8d282840546c02831e0d529865f225eebbc9327e95cec44319a78566","observation_id":"074e755d-8d82-4cfc-9609-5b9462d03de3","resolution":{"observed_at":"2026-08-04T19:46:51.120204Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.280380Z","title":"Robust distortion-free watermarks for language models,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":24,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.280380Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:49d076fdd1931b90ac89adc21870168b864f483bb7e12e5c7f72e62ad7458db2","observation_id":"219bb5de-11c5-4d2e-a0ab-b65e2e9f7158","resolution":{"observed_at":"2026-08-04T19:46:51.280380Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.472649Z","title":"Undetectable watermarks for language models,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":25,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.472649Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:972ef7ef19fdb9cd2d1e102b3b8d8e5052949100aa36155474b15704c04dfe63","observation_id":"5a130255-fa94-4c60-aada-778758f4a513","resolution":{"observed_at":"2026-08-04T19:46:51.472649Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.475398Z","title":"Bimark: Unbiased multilayer watermarking for large language models,","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":26,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.475398Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:b577c792449be771ce4549572312b92c8c6ab3b778598305c41606337aac4977","observation_id":"711f4715-12be-4419-ae98-4a1b7e30a767","resolution":{"observed_at":"2026-08-04T19:46:51.475398Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2409.19708","last_updated":"2024-09-29T13:51:15Z","snapshot_observed_at":"2026-07-06T19:24:08.391909Z","submitted_at":"2024-09-29T13:51:15Z","title":"A Certified Robust Watermark For Large Language Models","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2409.19708","snapshot_observed_at":"2026-08-04T19:46:51.478031Z","title":"A certified robust watermark for large language models,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":27,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.478031Z"},"links":{"cited_paper":"/paper/2409.19708","citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:44e71c574554c49656d2b9dc7ed5bcb1ffe2e1f59e3671cbe495587723e7715c","observation_id":"d96a3aa7-4cf8-4296-b2e5-c40baf2907a1","resolution":{"observed_at":"2026-08-04T19:46:51.478031Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.481037Z","title":"Bypassing LLM watermarks with color- aware substitutions,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":28,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.481037Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:59dd22a8d7a701d65dc9a5dcafb6555d7968bce7229a344cbbb5ca41b86165c4","observation_id":"375488aa-3132-453d-8623-a8c37b759485","resolution":{"observed_at":"2026-08-04T19:46:51.481037Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.483600Z","title":"b 4: A black-box scrubbing attack on LLM watermarks,","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":29,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.483600Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:e04b1a3e8cf65068dbfda1303cc2deb24fa632f5b776c26aec971eddd1c4fc79","observation_id":"6a8047f6-5637-484d-860f-005adf515442","resolution":{"observed_at":"2026-08-04T19:46:51.483600Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.486079Z","title":"De-mark: Watermark removal in large language models,","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":30,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.486079Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:035bb0cd69a9d606e11c512c0864909e081896a28aae1751af2b7ae87f26f9c9","observation_id":"2e6c8b32-61c5-48a6-86b2-fb5f686054e0","resolution":{"observed_at":"2026-08-04T19:46:51.486079Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2407.14206","last_updated":"2025-02-05T09:13:15Z","snapshot_observed_at":"2026-07-06T18:48:58.038753Z","submitted_at":"2024-07-19T11:04:54Z","title":"Watermark Smoothing Attacks against Language Models","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2407.14206","snapshot_observed_at":"2026-08-04T19:46:51.488748Z","title":"Watermark smoothing attacks against language models,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":31,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.488748Z"},"links":{"cited_paper":"/paper/2407.14206","citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:ac167e8160b0ee50c32e00dab65ac912dbbaf88b1e892841afe5639aa3ae15e3","observation_id":"01dcd123-91e0-498c-b77f-1eee60e2b138","resolution":{"observed_at":"2026-08-04T19:46:51.488748Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.491415Z","title":"Paraphrasing evades detectors of ai-generated text, but retrieval is an effective defense,","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":32,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.491415Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:e0479a03094b960d1a32991ac58735169e5800f7a376eebd87cab99009627ae2","observation_id":"b47e50ba-84ee-4318-81ef-6f983f7a5180","resolution":{"observed_at":"2026-08-04T19:46:51.491415Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.493900Z","title":"On the reliability of watermarks for large language models,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":33,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.493900Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:aced99071290766a4230aa0c9a194c10eedc7a38336e58457fa3b7918902ac91","observation_id":"170903a3-188e-463e-b5f2-3ed59e5dcd1e","resolution":{"observed_at":"2026-08-04T19:46:51.493900Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.496467Z","title":"Markmywords: Analyzing and evaluating language model watermarks,","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":34,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.496467Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:2609db954c6586f46d249126bddb2fee7c04e2cfea9c89f506cf03f6a85ecaee","observation_id":"2bac9422-7bc1-44a2-a67e-2918d54d90a1","resolution":{"observed_at":"2026-08-04T19:46:51.496467Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.498900Z","title":"Textbugger: Generating adversarial text against real-world applications,","venue":null,"work_id":null,"year":2018},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":35,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.498900Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:438d46238bc0599105f51d6cc4f3d33ca47e13356925d6d6eb064bb331a52b21","observation_id":"d78e126a-9892-4403-baef-81335802e8ef","resolution":{"observed_at":"2026-08-04T19:46:51.498900Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.501431Z","title":"Textattack: A framework for adversarial attacks, data augmentation, and adversarial training in nlp,","venue":null,"work_id":null,"year":2020},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":36,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.501431Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:7144d5f77fa3376a5ea7958c921dd90a354ecd64544cc09bd407b60f5378a9c2","observation_id":"ab842ec1-e927-44b5-b1b1-6832a794a607","resolution":{"observed_at":"2026-08-04T19:46:51.501431Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.503983Z","title":"Black-box generation of adversarial text sequences to evade deep learning classifiers,","venue":null,"work_id":null,"year":2018},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":37,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.503983Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:2145a15455f9352cc6e3f774ac6e79a8e2a42eae79f0b0a818dc46f0334b0cdc","observation_id":"6656c59e-9302-4c0c-a6cb-e3934b8040d7","resolution":{"observed_at":"2026-08-04T19:46:51.503983Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2504.08798","last_updated":"2025-04-08T14:10:57Z","snapshot_observed_at":"2026-08-10T00:32:52.433135Z","submitted_at":"2025-04-08T14:10:57Z","title":"Exploring Gradient-Guided Masked Language Model to Detect Textual Adversarial Attacks","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2504.08798","snapshot_observed_at":"2026-08-04T19:46:51.506675Z","title":"Exploring gradient-guided masked language model to detect textual adversarial attacks,","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":38,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.506675Z"},"links":{"cited_paper":"/paper/2504.08798","citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:800546bfadc269ce9b68c79ae67ddf5dbcf2077e7ab9c0ed1253db00ede19f77","observation_id":"d3bcf567-a4fc-46af-be4b-6aa044c64693","resolution":{"observed_at":"2026-08-04T19:46:51.506675Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.509818Z","title":"Masked language model based textual adversarial example detection,","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":39,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.509818Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:a5b2fe4d85ed05992125b019a0e9f91f4147d7d002c637b09c1d4b03b6c26a13","observation_id":"95f0f9bc-0bf2-42c1-beab-257e035e8796","resolution":{"observed_at":"2026-08-04T19:46:51.509818Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.512236Z","title":"Watermark stealing in large language models,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":40,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.512236Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:cb76ff7b94799f82d6bafe24e6404d15a94349490fdd5a43901116e39005d00f","observation_id":"97c9dea4-216b-488f-9676-6b630a7290d5","resolution":{"observed_at":"2026-08-04T19:46:51.512236Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.514650Z","title":"Stealing watermarks of large language models via mixed integer programming,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":41,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.514650Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:1cb6357f437880ae091771c702c63d2f99c5bb1a4ed93e5cdc564eca275ec727","observation_id":"d655e9cd-ac77-41a0-a642-16ad44da3855","resolution":{"observed_at":"2026-08-04T19:46:51.514650Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.517041Z","title":"Can watermarks survive translation? on the cross-lingual consistency of text watermark for large language models,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":42,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.517041Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:0ebb60b507035fda298f9bf4c33174a1206a0368b3df1bdd611b2c29d96747ad","observation_id":"761aaf7b-3953-4a74-995b-dea8f5b81598","resolution":{"observed_at":"2026-08-04T19:46:51.517041Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2407.04794","last_updated":"2024-11-28T11:28:39Z","snapshot_observed_at":"2026-08-06T01:13:47.988333Z","submitted_at":"2024-07-05T18:09:06Z","title":"On Evaluating The Performance of Watermarked Machine-Generated Texts Under Adversarial Attacks","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2407.04794","snapshot_observed_at":"2026-08-04T19:46:51.519469Z","title":"On evaluating the performance of watermarked machine-generated texts under adversarial attacks,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":43,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.519469Z"},"links":{"cited_paper":"/paper/2407.04794","citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:a3a492497663f0a24a25b6546a52796cff175774ef6f97ad13682334a58ee655","observation_id":"a9f7d773-9aae-4b4c-a6a9-a8e97877789a","resolution":{"observed_at":"2026-08-04T19:46:51.519469Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.522573Z","title":"Silverspeak: Evading ai-generated text detectors using homoglyphs,","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":44,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.522573Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:7d77db87be879613d9d20d2e47efd9cc9f36625b81cc9a22f65bb137b759c6ed","observation_id":"f3ff928d-3c0c-4544-bc3b-eb3e2a033914","resolution":{"observed_at":"2026-08-04T19:46:51.522573Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.525081Z","title":"Watermarks in the sand: impossibility of strong watermarking for language models,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":45,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.525081Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:bf626c04dccfa7dfa2ff25d7989d1e2a6450bfa333450535268bf65be1fec8f7","observation_id":"83e67408-48f4-47d0-9c06-5f3d00c9a010","resolution":{"observed_at":"2026-08-04T19:46:51.525081Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.527481Z","title":"Watermark under fire: A robustness evaluation of llm watermarking,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":46,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.527481Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:f9c6e31bc010b932cd3acb40cd42f1dc270d660372a656a4a3382bb7752d31b7","observation_id":"446f6627-c982-4311-a849-1648856c03e0","resolution":{"observed_at":"2026-08-04T19:46:51.527481Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.529999Z","title":"Sslguard: A watermarking scheme for self-supervised learning pre-trained encoders,","venue":null,"work_id":null,"year":2022},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":47,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.529999Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:32c1d60eea3d7a61fd82ea11f79de8084126adb76c7fd38ca152f5a64f4f17ec","observation_id":"82c4f5a0-2b51-4e6b-835b-e0196476444f","resolution":{"observed_at":"2026-08-04T19:46:51.529999Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.532316Z","title":"Entangled watermarks as a defense against model extraction,","venue":null,"work_id":null,"year":1937},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":48,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.532316Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:4a02761cd94f6477977288b9cbb6327d98fc38c143f159c7702fdd9a582c886b","observation_id":"7ba6361f-37c2-4ab9-861b-aa5a11034f7d","resolution":{"observed_at":"2026-08-04T19:46:51.532316Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.534766Z","title":"Exploring the limits of transfer learning with a unified text-to-text transformer,","venue":null,"work_id":null,"year":2020},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":49,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.534766Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:d5ceef7915f1dceed5a0b520b36d468356c3a6e8f1a625393221a56d6d732d79","observation_id":"a518a146-2387-42d3-ad09-49313477f8c5","resolution":{"observed_at":"2026-08-04T19:46:51.534766Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.537247Z","title":"Synthid","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":50,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.537247Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:17d8a548865faf9ef1083fe8c69f0379b354f937b76ec357424540699bac18a0","observation_id":"2c7283d7-325b-49e1-ac60-945dc8da4060","resolution":{"observed_at":"2026-08-04T19:46:51.537247Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.539643Z","title":"Terms of use,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":51,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.539643Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:9fa602a00d2ad8bdeec1b8f85ce684a04daab3e710e21141b8c99bb34ba16c32","observation_id":"7c481b3a-6f21-4141-8a4a-20099744aeca","resolution":{"observed_at":"2026-08-04T19:46:51.539643Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.542218Z","title":"Optimizing adaptive attacks against content watermarks for language models,","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":52,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.542218Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:be55dbe72a1a3b8c4e2324ea8362225180647a9b33decdc372d9e6d0ca9e3677","observation_id":"888b75bd-7e67-4ff7-a710-539909486d53","resolution":{"observed_at":"2026-08-04T19:46:51.542218Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.544695Z","title":"Markllm: An open-source toolkit for llm watermarking,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":53,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.544695Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:eccc346314df1e8fd3989268cf992d6c98ac1c6da93439505c0cc49737e60c35","observation_id":"532e6623-8cfb-4399-8b10-9d69ce010d5c","resolution":{"observed_at":"2026-08-04T19:46:51.544695Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.547241Z","title":"Software Framework for Topic Modelling with Large Corpora,","venue":null,"work_id":null,"year":2010},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":54,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.547241Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:e9095da4979878b782aab96fb392f683ca2b876abd988b73abb9f91fec8ef517","observation_id":"5ebde10f-8818-4b65-9628-7e3fc2812608","resolution":{"observed_at":"2026-08-04T19:46:51.547241Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2503.08716","last_updated":"2025-03-10T12:41:05Z","snapshot_observed_at":"2026-08-10T03:15:41.275566Z","submitted_at":"2025-03-10T12:41:05Z","title":"AuthorMist: Evading AI Text Detectors with Reinforcement Learning","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2503.08716","snapshot_observed_at":"2026-08-04T19:46:51.549738Z","title":"Authormist: Evading ai text detectors with reinforcement learning,","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":55,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.549738Z"},"links":{"cited_paper":"/paper/2503.08716","citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:5aa1b7a70421504f5874ff610c2a678bb7fb79aa6484ee7a471d58b6f80c3599","observation_id":"8079ca2f-7778-4935-8656-362c6ccf748a","resolution":{"observed_at":"2026-08-04T19:46:51.549738Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.552722Z","title":"Bleu: a method for automatic evaluation of machine translation,","venue":null,"work_id":null,"year":2002},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":56,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.552722Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:38e8e5a8987175bafa1a10e64bd7360f403de81b2a1954e97ee9e49f863049a3","observation_id":"17d1b733-e8c0-4662-88c5-e027750bdcd4","resolution":{"observed_at":"2026-08-04T19:46:51.552722Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.555135Z","title":"Rouge: A package for automatic evaluation of summaries,","venue":null,"work_id":null,"year":2004},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":57,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.555135Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:1ced6f368a2aa23662974367c6f2d513b273d02e21513c86eb6b99de1e19a7be","observation_id":"ad71345a-99b2-4dda-a860-b192500809ba","resolution":{"observed_at":"2026-08-04T19:46:51.555135Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.557552Z","title":"The cefr levels","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":58,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.557552Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:a32be5864c88ee529ef410fb93a2afc4a1d58d346a8349e8c026628f5febb339","observation_id":"c9135b88-33aa-44c1-9676-f03f0eb886ba","resolution":{"observed_at":"2026-08-04T19:46:51.557552Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.560050Z","title":"No free lunch in LLM watermarking: Trade-offs in watermarking design choices,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":59,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.560050Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:4e2d4327c8a060ee9f0063be4c3b1d79676d43c9a22befbea5aec7a7f28df5a6","observation_id":"2ded87d2-d40f-494a-a90e-7e672deb6a0b","resolution":{"observed_at":"2026-08-04T19:46:51.560050Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.562553Z","title":"New ai classifier for indicating ai-written tex,","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":60,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.562553Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:3ae8bde130b3659ec4631ac6bded81b491b275021b9219bc10d377d2544200d1","observation_id":"36368584-233b-4e00-9b85-40d1fc8c28aa","resolution":{"observed_at":"2026-08-04T19:46:51.562553Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.565140Z","title":"Ai detector by grammarly","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":61,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.565140Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:1e71397d2960b0979396010ea8d1dd647fa54a170ebe34014b3e69f7234ba62e","observation_id":"ac015a89-3fbc-4f7b-8369-27e998cde758","resolution":{"observed_at":"2026-08-04T19:46:51.565140Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.567775Z","title":"Gptzero’s ai detection technology","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":62,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.567775Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:0749c49dc895127c40a7107c1e78edb2914c55cc7c4f03b665151789bbf0bdb9","observation_id":"d492b1c6-5759-4b8c-9aef-ca60198a7e40","resolution":{"observed_at":"2026-08-04T19:46:51.567775Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.570265Z","title":"Bert: Pre-training of deep bidirectional transformers for language understanding,","venue":null,"work_id":null,"year":2019},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":63,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.570265Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:99f006c9ace6974f371baf9f5e0edacd12da85c52ec108b10610b3007139052f","observation_id":"bd70620d-e1db-403e-8409-964d3466b614","resolution":{"observed_at":"2026-08-04T19:46:51.570265Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.572662Z","title":"Languagetool","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":64,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.572662Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:a4140f74e7be1592958e457763bb405a9f5d3d9304413f817bf9fdd4da2fcf62","observation_id":"1852bfb4-a84b-401e-843f-68f7227f0ebf","resolution":{"observed_at":"2026-08-04T19:46:51.572662Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.575026Z","title":"Python tesseract","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":65,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.575026Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:b9b9e3a6aef05029d7bd100045e0bbe13042595b5e17c21c62b62bfc91111b3a","observation_id":"6635be85-c3ec-4b7c-b9f1-e6027b16a80e","resolution":{"observed_at":"2026-08-04T19:46:51.575026Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.577525Z","title":"Unicodedata","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":66,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.577525Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:6610f43715d3f5da79912e8c1caa881a55516cd986ccff0f2099172ae97a6d5e","observation_id":"94603794-456f-49da-ae57-f6817e1f2a8d","resolution":{"observed_at":"2026-08-04T19:46:51.577525Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.580631Z","title":"Results are reported forn= 5,10, and15, using both token- level and character-level perturbations across five watermark- ing schemes","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":67,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.580631Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:86b413361be67d17339017c300fd894c8c904ca94bd7686ae54b6960cad00780","observation_id":"27693111-82a8-4098-bf69-763669810bde","resolution":{"observed_at":"2026-08-04T19:46:51.580631Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.583360Z","title":"We compare both token-level (dashed lines) and character-level (solid lines) perturbations","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":68,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.583360Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:2a90f9634d97c4eb12da26c716a9066cc9c1e5cd82a7fe591fadb12bfd14665b","observation_id":"11a52c92-b1d3-4a3b-8614-fd49f5be26b2","resolution":{"observed_at":"2026-08-04T19:46:51.583360Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.586020Z","title":"As introduced TABLE XV: Performance of reference detectors with three levels of data augmentation across five watermark schemes","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":69,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.586020Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:7f278a0c937fdbd91d395cdb8390113ca7ea27a4b46ef634319e434667752d51","observation_id":"5b3f3546-5755-4d2d-8d3a-d65d2fbb1f25","resolution":{"observed_at":"2026-08-04T19:46:51.586020Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.588729Z","title":"Major Revision","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":70,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.588729Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:530887e7da93977644a7cac00221acbfc85a0b4ad1283d45c3183b257703d058","observation_id":"3a20e059-ac93-4eb6-958c-611393b97e47","resolution":{"observed_at":"2026-08-04T19:46:51.588729Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":"10.5281/zenodo.15872569","metadata_source":"openalex","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-05T02:28:24.338817Z","title":"We recom- mend downloading the code using the following command: git clone https://github.com/plll4zzx/CharacterRemoval4WM","venue":"Zenodo (CERN European Organization for Nuclear Research)","work_id":"c9f1a37a-b299-4fca-aeed-9be38d102145","year":null},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":71,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.591359Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:724a61b555bdafeee2e4775e07609458b64c3365bf3335780b1a1d383bfdec45","observation_id":"1b166493-c74c-4c53-9781-266e1cd21bfc","resolution":{"observed_at":"2026-08-04T19:49:01.231791Z","resolver_source":"doi","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-09T06:31:02.800959+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-09T06:31:02.800959+00:00","source":"crossref"},{"observed_at":"2026-08-09T06:30:57.326959+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.594038Z","title":null,"venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":72,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.594038Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:ae356cc9a5c602d2a2ef69a29a9cd18fef5648c69645553aa43e116c54a5a269","observation_id":"61a40ad6-0521-4c50-9d3e-e998b30e21ea","resolution":{"observed_at":"2026-08-04T19:46:51.594038Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.596929Z","title":null,"venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":73,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.596929Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:797b28d40c6c0b0aa93eaaec33d1a5b6bf2493beb130e1e4528380e24fbbce3a","observation_id":"be7ba321-042e-441d-8d9f-2a6d5dd1cb3d","resolution":{"observed_at":"2026-08-04T19:46:51.596929Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.599453Z","title":"real- newslike/*","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":74,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.599453Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:547377a20ca4f6b7dbeff4e4e355cccb31076bdd6ffe5ffaa5458da4cab04e4c","observation_id":"602d3fd1-93ae-49ac-9d77-b6519a501039","resolution":{"observed_at":"2026-08-04T19:46:51.599453Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.602058Z","title":"/../dataset/c4/realnewslike","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":75,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.602058Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:bebd89ebe83e672c792e20e4b63bad50e8fca1dea88370d58061ad168d0ccd57","observation_id":"5c90de35-de9d-4060-8c8b-694dcd33c20b","resolution":{"observed_at":"2026-08-04T19:46:51.602058Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-04T19:46:51.604848Z","title":"KGW\" --num_epochs 15 \\ --rand_char_rate 0.15 --rand_times 9 \\ --llm_name","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks","version":2},"reference_index":76,"source":"pdf_text","source_observed_at":"2026-08-04T19:46:51.604848Z"},"links":{"citing_paper":"/paper/2509.09112"},"observation_digest":"sha256:6dcbb83a5748521ad99d2262411b2078b0dbc41c638a43232c9ca997205886b8","observation_id":"0dc88317-ae32-4354-b1ec-0980abe5b17e","resolution":{"observed_at":"2026-08-04T19:46:51.604848Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}}],"paper":{"arxiv_id":"2509.09112","last_updated":"2025-09-14T07:46:04Z","latest_version":2,"primary_category":"cs.CR","snapshot_observed_at":"2026-08-06T21:57:01.794591Z","submitted_at":"2025-09-11T02:50:07Z","title":"Character-Level Perturbations Disrupt LLM Watermarks"},"reference_resolution":{"displayed":76,"state_counts":{"malformed_identifier":0,"metadata_mismatch":0,"parse_uncertain":0,"unresolved":75,"verified_exact":1,"verified_fuzzy":0},"total_outbound_references":76},"refusal":"A citation records a reference. It does not transfer a finding from one paper to another.","schema":"pith.paper-citation-record.v1","standing_sources":[{"observed_at":"2026-08-09T06:31:02.800959+00:00","source":"crossref"},{"observed_at":"2026-08-09T06:30:57.326959+00:00","source":"retraction_watch"}],"thesis":"As of 10 August 2026, this Paper Citation Record lists 76 of 76 outbound references and 1 inbound Pith citation observation for arXiv:2509.09112."}