{"as_of":"2026-08-08T23:12:00Z","caps":{"database_statements":6,"inbound":100,"outbound":100},"context_digest":"sha256:8af72ec85f0aa8c44997ddd41ad06fe294d597a3861fef720023bdd9ee0c5c35","coverage":[{"denominator":22,"lane":"reference_resolution","note":"Typed states for the displayed outbound observations.","records_observed":22,"source":"paper_references, paper_reference_links","source_observed_at":"2026-08-02T22:51:31.497576Z","state":"measured"},{"denominator":22,"lane":"standing_notices","note":"One-hop event checks from named stored sources.","records_observed":22,"source":"scholarly_work_events, retraction_status_cache","source_observed_at":"2026-08-08T06:32:00.761636+00:00","state":"measured"},{"denominator":0,"lane":"inbound_itemization","note":"Pith citing papers itemized under the disclosed page cap.","records_observed":0,"source":"paper_references, paper_reference_links","source_observed_at":null,"state":"measured"},{"denominator":1,"lane":"external_citation_measurements","note":"A source-named dated measurement, never combined with another source.","records_observed":0,"source":"cited_works","source_observed_at":null,"state":"measured"}],"external_citation_measurements":[],"inbound":[],"links":{"evidence":"/evidence","html":"/paper/2602.16065/citation-record","integrity":"/paper/2602.16065/integrity","json":"/paper/2602.16065/citation-record.json","paper":"/paper/2602.16065"},"outbound":[{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-02T22:51:31.417228Z","title":null,"venue":null,"work_id":null,"year":1976},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":1,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:31.417228Z"},"links":{"citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:a5235e828bb6aa529d070c048bec3e8e47d3ee59e82915c233046b099acb94c8","observation_id":"cee95578-7f1b-4f8d-bed9-4b5d3e1c6dc9","resolution":{"observed_at":"2026-08-02T22:51:31.417228Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.08044","last_updated":"2024-10-10T15:36:10Z","snapshot_observed_at":"2026-08-08T12:23:33.512681Z","submitted_at":"2024-10-10T15:36:10Z","title":"The Rise of AI-Generated Content in Wikipedia","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.08044","snapshot_observed_at":"2026-08-02T22:51:29.355402Z","title":"The rise of ai-generated content in wikipedia.arXiv preprint arXiv:2410.08044,","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":3,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:29.355402Z"},"links":{"cited_paper":"/paper/2410.08044","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:8d706a58ed6200e337489e341777e6d3be2af546bbf987716938f570befcc23c","observation_id":"3fee4ba4-1f9a-4785-9009-539be2e30a9b","resolution":{"observed_at":"2026-08-02T22:51:29.355402Z","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":"2026-08-02T22:51:29.612552Z","title":"Accessed: 2025-06-26","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":5,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:29.612552Z"},"links":{"citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:fd05527d2cf89a716182a84bc25e7403fb93f5247e316dfc67ddcdcb965ec85c","observation_id":"56de18f0-163f-4ff7-ab8d-180ab2c89b04","resolution":{"observed_at":"2026-08-02T22:51:29.612552Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"1710.07535","last_updated":"2017-11-23T16:28:48Z","snapshot_observed_at":"2026-08-03T17:45:35.865535Z","submitted_at":"2017-10-19T16:04:05Z","title":"Data-Free Knowledge Distillation for Deep Neural Networks","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1710.07535","snapshot_observed_at":"2026-08-02T22:51:30.218317Z","title":"Data-free knowledge distillation for deep neural networks.arXiv preprint arXiv:1710.07535,","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":9,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:30.218317Z"},"links":{"cited_paper":"/paper/1710.07535","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:27ae6ce9e59cbbb7595ab25c00706aa514e9a428005a11a652da04cb913471a7","observation_id":"0abc1751-955e-409b-bf97-d44f4cf30030","resolution":{"observed_at":"2026-08-02T22:51:30.218317Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"1610.05755","last_updated":"2017-03-03T18:56:43Z","snapshot_observed_at":"2026-07-06T05:15:05.941444Z","submitted_at":"2016-10-18T19:37:37Z","title":"Semi-supervised Knowledge Transfer for Deep Learning from Private Training Data","version":4},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1610.05755","snapshot_observed_at":"2026-08-02T22:51:30.546228Z","title":"Semi-supervised knowledge transfer for deep learning from private training data.arXiv preprint arXiv:1610.05755,","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":11,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:30.546228Z"},"links":{"cited_paper":"/paper/1610.05755","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:46ab9ae99fb7e363a53e3082be516e3f91cff04a21be3d0988bfa6194da8b4f8","observation_id":"87ac8004-9039-444e-ad09-754fff7b8bd9","resolution":{"observed_at":"2026-08-02T22:51:30.546228Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.18779","last_updated":"2024-10-24T14:31:52Z","snapshot_observed_at":"2026-07-06T19:39:06.300432Z","submitted_at":"2024-10-24T14:31:52Z","title":"A Little Help Goes a Long Way: Efficient LLM Training by Leveraging Small LMs","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.18779","snapshot_observed_at":"2026-08-02T22:51:30.658443Z","title":"A little help goes a long way: Efficient llm training by leveraging small lms.arXiv preprint arXiv:2410.18779,","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":12,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:30.658443Z"},"links":{"cited_paper":"/paper/2410.18779","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:f0d90adfc2dfe940cac8e8cedbeb672782d3202d8f589503cfe3e7cd825a5383","observation_id":"986e5630-acd8-489d-88ee-9c0e0a79ee7b","resolution":{"observed_at":"2026-08-02T22:51:30.658443Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2305.17493","last_updated":"2024-04-14T05:20:10Z","snapshot_observed_at":"2026-08-05T16:03:40.517679Z","submitted_at":"2023-05-27T15:10:41Z","title":"The Curse of Recursion: Training on Generated Data Makes Models Forget","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2305.17493","snapshot_observed_at":"2026-08-02T22:51:31.062816Z","title":"The curse of recursion: Training on generated data makes models forget","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":16,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:31.062816Z"},"links":{"cited_paper":"/paper/2305.17493","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:8c30cf48372d5e56a36f0d59c3e207e5d0bf0ea98b28c264c9a19a6bd8165906","observation_id":"ed225bc6-5d12-40b8-bf3d-76e47ddcb511","resolution":{"observed_at":"2026-08-02T22:51:31.062816Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2412.17646","last_updated":"2024-12-23T15:21:50Z","snapshot_observed_at":"2026-07-06T20:12:13.548098Z","submitted_at":"2024-12-23T15:21:50Z","title":"Rate of Model Collapse in Recursive Training","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2412.17646","snapshot_observed_at":"2026-08-02T22:51:31.140572Z","title":"Rate of model collapse in recursive training.arXiv preprint arXiv:2412.17646,","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":17,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:31.140572Z"},"links":{"cited_paper":"/paper/2412.17646","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:46eb5cabeb2a9c2e1eb668ab8afdf28ba18641fff64bdcb3486e5bc8e42e434e","observation_id":"855e7720-495d-4353-bca4-bbb0968c09e4","resolution":{"observed_at":"2026-08-02T22:51:31.140572Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2409.20124","last_updated":"2024-09-30T09:25:27Z","snapshot_observed_at":"2026-08-05T18:17:21.871046Z","submitted_at":"2024-09-30T09:25:27Z","title":"Conditional Diffusion Models are Minimax-Optimal and Manifold-Adaptive for Conditional Distribution Estimation","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2409.20124","snapshot_observed_at":"2026-08-02T22:51:31.190156Z","title":"Conditional diffusion models are minimax- optimal and manifold-adaptive for conditional distribution estimation.arXiv preprint arXiv:2409.20124,","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":18,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:31.190156Z"},"links":{"cited_paper":"/paper/2409.20124","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:ccedbf18848e588618218c3cbfcd4999fd8f17ca664bd348a8d585e2f28fc974","observation_id":"8ce66746-1de4-401b-a93a-e901182230ae","resolution":{"observed_at":"2026-08-02T22:51:31.190156Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"1802.06739","last_updated":"2018-02-19T18:05:33Z","snapshot_observed_at":"2026-07-06T06:24:12.439943Z","submitted_at":"2018-02-19T18:05:33Z","title":"Differentially Private Generative Adversarial Network","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1802.06739","snapshot_observed_at":"2026-08-02T22:51:31.257678Z","title":"Differentially private generative adversarial network.arXiv preprint arXiv:1802.06739,","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":19,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:31.257678Z"},"links":{"cited_paper":"/paper/1802.06739","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:5f66a855151594cdc92162d1174502b18fb0b8419bf724e5bcb7860f4dd1eec2","observation_id":"f6447816-1037-462d-92d2-2b269a1ced9e","resolution":{"observed_at":"2026-08-02T22:51:31.257678Z","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":"2026-08-02T22:51:31.497576Z","title":"At iterationt, a batch of m1 new samples fromP 0 is appended to the dataset, together withm 2 = ((1−α)/α)m 1 synthetic samples generated from the previous iterate bPt−1","venue":null,"work_id":null,"year":2021},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":22,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:31.497576Z"},"links":{"citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:cd3607b17adf24886cd4a3e300c59bb5ac396fd0fee5e20e380f125e6caf2a34","observation_id":"7d138cf6-941f-4cd0-aad0-2e387a0c9b19","resolution":{"observed_at":"2026-08-02T22:51:31.497576Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2404.01413","last_updated":"2024-04-29T23:13:42Z","snapshot_observed_at":"2026-07-06T17:54:11.726535Z","submitted_at":"2024-04-01T18:31:24Z","title":"Is Model Collapse Inevitable? Breaking the Curse of Recursion by Accumulating Real and Synthetic Data","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2404.01413","snapshot_observed_at":"2026-08-02T22:51:29.726873Z","title":"Is model collapse inevitable? breaking the curse of recursion by accumulating real and synthetic data.arXiv preprint arXiv:2404.01413,","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":1959,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:29.726873Z"},"links":{"cited_paper":"/paper/2404.01413","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:c660f62611db0e9cf2223e4feca6911891420bd20a362411734e99cc6856e97e","observation_id":"c0df71f4-7fe1-46a5-8290-a8ed49364dab","resolution":{"observed_at":"2026-08-02T22:51:29.726873Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"1611.04201","last_updated":"2017-06-08T07:21:39Z","snapshot_observed_at":"2026-07-06T05:18:28.057975Z","submitted_at":"2016-11-13T23:08:42Z","title":"CAD2RL: Real Single-Image Flight without a Single Real Image","version":4},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1611.04201","snapshot_observed_at":"2026-08-02T22:51:30.768635Z","title":"Cad2rl: Real single-image flight without a single real image.arXiv preprint arXiv:1611.04201,","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":1976,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:30.768635Z"},"links":{"cited_paper":"/paper/1611.04201","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:11f837f0c2b6241d69c06e96694fedc785924aa0227c4b5cea8cd9ed795b3cf8","observation_id":"c628425b-bc60-48dc-931f-1ccdb861e2f1","resolution":{"observed_at":"2026-08-02T22:51:30.768635Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"1503.02531","last_updated":"2015-03-09T15:44:49Z","snapshot_observed_at":"2026-07-06T04:11:24.157003Z","submitted_at":"2015-03-09T15:44:49Z","title":"Distilling the Knowledge in a Neural Network","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1503.02531","snapshot_observed_at":"2026-08-02T22:51:29.886014Z","title":"Distilling the knowledge in a neural network","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":2009,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:29.886014Z"},"links":{"cited_paper":"/paper/1503.02531","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:a1e58fbbc01156b4f6c4e683bb7fb1d466cb451c8a1e9cc474cacb24a8aaf190","observation_id":"d99ac721-78bd-4634-adcf-a95836fbf90f","resolution":{"observed_at":"2026-08-02T22:51:29.886014Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"1910.01108","last_updated":"2020-03-01T02:57:50Z","snapshot_observed_at":"2026-08-07T19:07:36.327251Z","submitted_at":"2019-10-02T17:56:28Z","title":"DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter","version":4},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1910.01108","snapshot_observed_at":"2026-08-02T22:51:30.869458Z","title":"Distilbert, a distilled version of bert: smaller, faster, cheaper and lighter.arXiv preprint arXiv:1910.01108,","venue":null,"work_id":null,"year":1910},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":2016,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:30.869458Z"},"links":{"cited_paper":"/paper/1910.01108","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:535007a205c23c9740a1c22adbeb71a78dab1468443d144e5e658c55478ce56a","observation_id":"8646fdf0-d8e9-4f19-898f-453c900f48a5","resolution":{"observed_at":"2026-08-02T22:51:30.869458Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"1712.08244","last_updated":"2018-02-16T21:19:40Z","snapshot_observed_at":"2026-08-05T12:02:35.070730Z","submitted_at":"2017-12-21T23:13:27Z","title":"How Well Can Generative Adversarial Networks Learn Densities: A Nonparametric View","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1712.08244","snapshot_observed_at":"2026-08-02T22:51:30.051221Z","title":"Pseudo-label: The simple and efficient semi-supervised learning method for deep neural networks","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":2018,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:30.051221Z"},"links":{"cited_paper":"/paper/1712.08244","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:4b25e3711865f361265629e9b552b8037607ad7c9a61cb783fa03704dbfe1474","observation_id":"2f782262-4796-4697-a9b9-7f08470d77ee","resolution":{"observed_at":"2026-08-02T22:51:30.051221Z","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":"2026-08-02T22:51:30.967040Z","title":"The woman worked as a babysitter: On biases in language generation","venue":null,"work_id":null,"year":2019},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":2019,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:30.967040Z"},"links":{"citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:1e716fd9004ea50d4d5a21cd6de037068756ad3a79007fbd48aa74332edd5257","observation_id":"35a2193f-707a-4e97-95d9-c876e61555c1","resolution":{"observed_at":"2026-08-02T22:51:30.967040Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2403.02726","last_updated":"2024-03-05T07:34:41Z","snapshot_observed_at":"2026-08-08T18:03:38.031068Z","submitted_at":"2024-03-05T07:34:41Z","title":"Bias in Generative AI","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2403.02726","snapshot_observed_at":"2026-08-02T22:51:31.337608Z","title":"Bias in generative ai.arXiv preprint arXiv:2403.02726,","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":2020,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:31.337608Z"},"links":{"cited_paper":"/paper/2403.02726","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:de4e70fcc6091ac6406aa156a1a4abdfe8fe24b2825a098e161331e1c980937d","observation_id":"fea741e6-1c88-424c-9151-05bc83426b09","resolution":{"observed_at":"2026-08-02T22:51:31.337608Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2303.01255","last_updated":"2023-02-17T17:39:41Z","snapshot_observed_at":"2026-08-04T13:44:42.648781Z","submitted_at":"2023-02-17T17:39:41Z","title":"Combining Generative Artificial Intelligence (AI) and the Internet: Heading towards Evolution or Degradation?","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2303.01255","snapshot_observed_at":"2026-08-02T22:51:30.382113Z","title":"Combining generative artificial intelligence (ai) and the internet: Heading towards evolution or degradation?arXiv preprint arXiv:2303.01255,","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":2021,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:30.382113Z"},"links":{"cited_paper":"/paper/2303.01255","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:ed2b4621559a0003a5e5c7d88ce664d94791ff4418a907d019e2f5e63c87a401","observation_id":"d21097e9-9033-42da-9c89-432c69c68e6a","resolution":{"observed_at":"2026-08-02T22:51:30.382113Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"1810.10863","last_updated":"2018-10-25T13:17:33Z","snapshot_observed_at":"2026-08-07T09:09:39.188154Z","submitted_at":"2018-10-25T13:17:33Z","title":"GAN Augmentation: Augmenting Training Data using Generative Adversarial Networks","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1810.10863","snapshot_observed_at":"2026-08-02T22:51:29.348864Z","title":"Gan augmentation: Augmenting training data using generative adver- sarial networks.arXiv preprint arXiv:1810.10863,","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":2023,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:29.348864Z"},"links":{"cited_paper":"/paper/1810.10863","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:930e6298ae67fc987a57db7d81dcb935774804a39c560056582bc01ab85cf15c","observation_id":"938a8f11-b6cc-4d76-a3d5-4585227e8376","resolution":{"observed_at":"2026-08-02T22:51:29.348864Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2208.05314","last_updated":"2023-05-29T11:12:13Z","snapshot_observed_at":"2026-08-06T13:23:45.016489Z","submitted_at":"2022-08-10T12:50:47Z","title":"Convergence of denoising diffusion models under the manifold hypothesis","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2208.05314","snapshot_observed_at":"2026-08-02T22:51:29.458993Z","title":"Convergence of denoising diffusion models under the manifold hypoth- esis.arXiv preprint arXiv:2208.05314,","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":2024,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:29.458993Z"},"links":{"cited_paper":"/paper/2208.05314","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:6aecb7f675e129698a6f147bdee76d705aef36474b00506bfdec5b20b40e1c6c","observation_id":"3391c67b-6c9c-466e-a363-954df390cd00","resolution":{"observed_at":"2026-08-02T22:51:29.458993Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2310.00429","last_updated":"2024-04-02T14:09:40Z","snapshot_observed_at":"2026-07-06T16:25:58.486204Z","submitted_at":"2023-09-30T16:41:04Z","title":"On the Stability of Iterative Retraining of Generative Models on their own Data","version":5},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2310.00429","snapshot_observed_at":"2026-08-02T22:51:29.223011Z","title":"On the stability of iterative retraining of generative models on their own data.arXiv preprint arXiv:2310.00429,","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training","version":2},"reference_index":2025,"source":"pdf_text","source_observed_at":"2026-08-02T22:51:29.223011Z"},"links":{"cited_paper":"/paper/2310.00429","citing_paper":"/paper/2602.16065"},"observation_digest":"sha256:9af6e642871f58c55bf25d816c145bf15e2be39fd0c33b7bc50b0e87513d42a7","observation_id":"10be65be-fe0c-4977-a48b-fb97efcbe485","resolution":{"observed_at":"2026-08-02T22:51:29.223011Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}}],"paper":{"arxiv_id":"2602.16065","last_updated":"2026-06-26T00:44:59Z","latest_version":2,"primary_category":"cs.LG","snapshot_observed_at":"2026-08-08T10:24:22.445959Z","submitted_at":"2026-02-17T22:38:18Z","title":"Can Generative Artificial Intelligence Survive Data Contamination? Theoretical Guarantees under Contaminated Recursive Training"},"reference_resolution":{"displayed":22,"state_counts":{"malformed_identifier":0,"metadata_mismatch":0,"parse_uncertain":0,"unresolved":22,"verified_exact":0,"verified_fuzzy":0},"total_outbound_references":22},"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":"2026-08-08T06:32:00.761636+00:00","source":"crossref"},{"observed_at":"2026-08-08T06:31:55.24221+00:00","source":"retraction_watch"}],"thesis":"As of 8 August 2026, this Paper Citation Record lists 22 of 22 outbound references and 0 inbound Pith citation observations for arXiv:2602.16065."}