{"as_of":"2026-08-19T04:36:00Z","caps":{"database_statements":6,"inbound":100,"outbound":100},"context_digest":"sha256:5a04bc4f7b4a5ce09e3bba78bbd433f52715471d68af1deb100dbffbb877a947","coverage":[{"denominator":20,"lane":"reference_resolution","note":"Typed states for the displayed outbound observations.","records_observed":20,"source":"paper_references, paper_reference_links","source_observed_at":"2026-08-16T06:00:23.952403Z","state":"measured"},{"denominator":20,"lane":"standing_notices","note":"One-hop event checks from named stored sources.","records_observed":20,"source":"scholarly_work_events, retraction_status_cache","source_observed_at":"2026-08-18T06:34:40.430872+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/2504.19333/citation-record","integrity":"/paper/2504.19333/integrity","json":"/paper/2504.19333/citation-record.json","paper":"/paper/2504.19333"},"outbound":[{"citation":{"cited_paper":{"arxiv_id":"2312.06795","last_updated":"2024-08-10T00:02:00Z","snapshot_observed_at":"2026-08-18T21:12:09.108009Z","submitted_at":"2023-12-11T19:10:55Z","title":"Model Breadcrumbs: Scaling Multi-Task Model Merging with Sparse Masks","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2312.06795","snapshot_observed_at":"2026-08-16T06:00:23.862503Z","title":"Model breadcrumbs: Scaling multi-task model merging with sparse masks","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":2,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.862503Z"},"links":{"cited_paper":"/paper/2312.06795","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:739db0e43cc03bc2cbb9a7e8e1a951215255ac02a00355f5997384944efb26ea","observation_id":"3df58c38-8eb6-48f4-bf73-08e9682129a5","resolution":{"observed_at":"2026-08-16T06:00:23.862503Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2201.06025","last_updated":"2022-10-19T07:59:03Z","snapshot_observed_at":"2026-08-19T01:16:06.703282Z","submitted_at":"2022-01-16T11:47:23Z","title":"COLD: A Benchmark for Chinese Offensive Language Detection","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2201.06025","snapshot_observed_at":"2026-08-16T06:00:23.867874Z","title":"Cold: A benchmark for chinese offensive language detection","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":3,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.867874Z"},"links":{"cited_paper":"/paper/2201.06025","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:4e6a0d9ab0a0985eff2fb2c43c2ff6336d9303ea1d65805d0529424f6870e174","observation_id":"fa8752b8-dae6-4fad-93aa-06a684648f2b","resolution":{"observed_at":"2026-08-16T06:00:23.867874Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2212.04089","last_updated":"2023-03-31T15:27:01Z","snapshot_observed_at":"2026-08-13T03:27:01.609831Z","submitted_at":"2022-12-08T05:50:53Z","title":"Editing Models with Task Arithmetic","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2212.04089","snapshot_observed_at":"2026-08-16T06:00:23.883875Z","title":"Editing models with task arithmetic","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":6,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.883875Z"},"links":{"cited_paper":"/paper/2212.04089","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:0e095f0d24ac6b3682f8826622e861339d39973682ed24a5eff0d55cce5aef09","observation_id":"32443bf3-e43f-427a-b4ac-680816f81556","resolution":{"observed_at":"2026-08-16T06:00:23.883875Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2312.06674","last_updated":"2023-12-07T19:40:50Z","snapshot_observed_at":"2026-08-14T15:42:19.849118Z","submitted_at":"2023-12-07T19:40:50Z","title":"Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversations","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2312.06674","snapshot_observed_at":"2026-08-16T06:00:23.889544Z","title":"Llama guard: Llm-based input-output safeguard for human-ai conversations","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":7,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.889544Z"},"links":{"cited_paper":"/paper/2312.06674","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:3af9a0312b3714117783b98fcf59dc12788d4f9ef00a086123d08b42b6edfdd3","observation_id":"a36b388d-c54a-40c3-8e6f-fb8ce057865f","resolution":{"observed_at":"2026-08-16T06:00:23.889544Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2311.12786","last_updated":"2024-08-21T16:37:20Z","snapshot_observed_at":"2026-08-19T01:15:11.526401Z","submitted_at":"2023-11-21T18:51:04Z","title":"Mechanistically analyzing the effects of fine-tuning on procedurally defined tasks","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2311.12786","snapshot_observed_at":"2026-08-16T06:00:23.894731Z","title":"Mechanistically analyzing the effects of fine-tuning on procedurally defined tasks","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":8,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.894731Z"},"links":{"cited_paper":"/paper/2311.12786","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:be4a080a60640e42e3d06f482f883a7bc4ff8a2ad151fbaa038923d74c601459","observation_id":"ccc7623c-1d06-4b4a-981f-d2ef55405bd1","resolution":{"observed_at":"2026-08-16T06:00:23.894731Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2212.09849","last_updated":"2025-05-21T23:53:00Z","snapshot_observed_at":"2026-08-16T16:07:30.940052Z","submitted_at":"2022-12-19T20:46:43Z","title":"Dataless Knowledge Fusion by Merging Weights of Language Models","version":6},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2212.09849","snapshot_observed_at":"2026-08-16T06:00:23.899382Z","title":"Dataless knowledge fusion by merging weights of language models","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":9,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.899382Z"},"links":{"cited_paper":"/paper/2212.09849","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:2c0883dc914935c0a310acccf97fb0385369984236660969d2ba7724bbb2cdd8","observation_id":"eade8bbf-d55c-435f-b5f4-2234a672d551","resolution":{"observed_at":"2026-08-16T06:00:23.899382Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2310.17389","last_updated":"2023-10-26T13:35:41Z","snapshot_observed_at":"2026-08-16T14:48:37.882921Z","submitted_at":"2023-10-26T13:35:41Z","title":"ToxicChat: Unveiling Hidden Challenges of Toxicity Detection in Real-World User-AI Conversation","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2310.17389","snapshot_observed_at":"2026-08-16T06:00:23.903994Z","title":"Toxicchat: Unveiling hidden challenges of toxicity detection in real-world user-ai conversation","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":10,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.903994Z"},"links":{"cited_paper":"/paper/2310.17389","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:6c7d7c5aae2d969c9c3429a09ea7adcbd4a3951c9cd5681906928145aeaf8172","observation_id":"dc7e57a3-1e0a-4af9-bf72-6eda56d6263f","resolution":{"observed_at":"2026-08-16T06:00:23.903994Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"1907.11692","last_updated":"2019-07-26T17:48:29Z","snapshot_observed_at":"2026-08-16T14:33:50.657682Z","submitted_at":"2019-07-26T17:48:29Z","title":"RoBERTa: A Robustly Optimized BERT Pretraining Approach","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1907.11692","snapshot_observed_at":"2026-08-16T06:00:23.908706Z","title":"Roberta: A robustly optimized bert pretraining approach","venue":null,"work_id":null,"year":1907},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":11,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.908706Z"},"links":{"cited_paper":"/paper/1907.11692","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:0d1089fa92f1964f611649f27001a7dfcbcc6f2ee6645b191eeadf1acabc2178","observation_id":"c260cfb5-bc3e-4afc-bc24-9a2b480215e5","resolution":{"observed_at":"2026-08-16T06:00:23.908706Z","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":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-16T06:00:24.269071Z","title":"Accessed: 2024-08-17","venue":null,"work_id":"c831b79c-26b7-4222-a581-f9f9112ac21a","year":2024},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":13,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.918277Z"},"links":{"citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:bebc1b0c43b7292060c06d74b4e85b70a8f386cdf79c9bfd2d92f6f2545a8651","observation_id":"def71c45-7082-45a1-b593-8e9428168a5f","resolution":{"observed_at":"2026-08-16T06:00:24.275479Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2303.08774","last_updated":"2024-03-04T06:01:33Z","snapshot_observed_at":"2026-08-17T09:58:46.058102Z","submitted_at":"2023-03-15T17:15:04Z","title":"GPT-4 Technical Report","version":6},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2303.08774","snapshot_observed_at":"2026-08-16T06:00:23.922780Z","title":"Gpt-4 technical report","venue":null,"work_id":null,"year":2021},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":14,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.922780Z"},"links":{"cited_paper":"/paper/2303.08774","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:e0d29197c9be1ae4f215dd5f9a9fbcd0bd934cae04fe84c009ce530900403a33","observation_id":"8ee93ba8-429e-4e69-b314-962e10bf93a9","resolution":{"observed_at":"2026-08-16T06:00:23.922780Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2310.10501","last_updated":"2023-10-16T15:20:30Z","snapshot_observed_at":"2026-08-16T14:51:42.822750Z","submitted_at":"2023-10-16T15:20:30Z","title":"NeMo Guardrails: A Toolkit for Controllable and Safe LLM Applications with Programmable Rails","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2310.10501","snapshot_observed_at":"2026-08-16T06:00:23.932513Z","title":"Nemo guardrails: A toolkit for controllable and safe llm applications with programmable rails","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":16,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.932513Z"},"links":{"cited_paper":"/paper/2310.10501","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:9733ba5fdf7743cca8cc9c7331412220ca2382cef30aac8b15a0cba8cb83b519","observation_id":"8b29e581-855c-4b23-8efc-cb5c7e31ec07","resolution":{"observed_at":"2026-08-16T06:00:23.932513Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2012.15761","last_updated":"2021-06-03T08:05:32Z","snapshot_observed_at":"2026-08-16T18:54:54.842746Z","submitted_at":"2020-12-31T17:36:48Z","title":"Learning from the Worst: Dynamically Generated Datasets to Improve Online Hate Detection","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2012.15761","snapshot_observed_at":"2026-08-16T06:00:23.937388Z","title":"Learning from the worst: Dynamically generated datasets to improve online hate detection.arXiv preprint arXiv:2012.15761,","venue":null,"work_id":null,"year":2012},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":17,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.937388Z"},"links":{"cited_paper":"/paper/2012.15761","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:92f6b5bff8939b7d04ccb54e11de9f81b7ae31dcf0e1f0241bb1c3210192d594","observation_id":"5265f050-71ce-40d1-8a2e-f6d252ca3b67","resolution":{"observed_at":"2026-08-16T06:00:23.937388Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"1609.04468","last_updated":"2016-12-06T14:39:05Z","snapshot_observed_at":"2026-08-17T08:48:57.346783Z","submitted_at":"2016-09-14T22:42:23Z","title":"Sampling Generative Networks","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"1609.04468","snapshot_observed_at":"2026-08-16T06:00:23.947594Z","title":"Sampling generative networks","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":19,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.947594Z"},"links":{"cited_paper":"/paper/1609.04468","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:08623c2643a2df53ef19a194b46777ed114c4af5c4ebf5d01b0c8108ff7e4a41","observation_id":"7f275c91-36c7-4d68-83a7-cec8d50ee039","resolution":{"observed_at":"2026-08-16T06:00:23.947594Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2311.13534","last_updated":"2023-12-08T16:45:22Z","snapshot_observed_at":"2026-08-16T14:41:05.977260Z","submitted_at":"2023-11-22T17:14:54Z","title":"LM-Cocktail: Resilient Tuning of Language Models via Model Merging","version":4},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2311.13534","snapshot_observed_at":"2026-08-16T06:00:23.952403Z","title":"Lm-cocktail: Resilient tuning of language models via model merging","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":20,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.952403Z"},"links":{"cited_paper":"/paper/2311.13534","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:fb6ac4b8a6e0b4030dc7e5f1cb09c956cbf7d8f0ee449bc197293f7c26e49b87","observation_id":"23735345-7d8c-4763-bd71-b93ed6809af0","resolution":{"observed_at":"2026-08-16T06:00:23.952403Z","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":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-16T06:00:24.286297Z","title":"Primeguard: Safe and helpful llms through tuning-free routing","venue":null,"work_id":"cfe529aa-1809-452a-989e-7416f531e79f","year":2024},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":2019,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.913632Z"},"links":{"citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:b609fa2320499ccbb6937c83fbc7eae0e27412e528454f1bc263506cebfe3003","observation_id":"3ff7dcd4-c924-4361-9baa-8819c010a11e","resolution":{"observed_at":"2026-08-16T06:00:24.291950Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2402.05672","last_updated":"2024-02-08T13:47:50Z","snapshot_observed_at":"2026-08-12T15:58:37.148545Z","submitted_at":"2024-02-08T13:47:50Z","title":"Multilingual E5 Text Embeddings: A Technical Report","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2402.05672","snapshot_observed_at":"2026-08-16T06:00:23.942772Z","title":"Multilin- gual e5 text embeddings: A technical report","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":2020,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.942772Z"},"links":{"cited_paper":"/paper/2402.05672","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:273fc14c1e20103df18145d8c68a8bdf5d74b78ab68b40bfeb68b2fd75b250ff","observation_id":"1e2d4b5c-4e60-4341-9a1c-7b87eda8b4bc","resolution":{"observed_at":"2026-08-16T06:00:23.942772Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2310.03693","last_updated":"2023-10-05T17:12:17Z","snapshot_observed_at":"2026-07-06T02:11:23.670680Z","submitted_at":"2023-10-05T17:12:17Z","title":"Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2310.03693","snapshot_observed_at":"2026-08-16T06:00:23.927806Z","title":"Fine-tuning aligned language models compromises safety, even when users do not intend to! arXiv preprint arXiv:2310.03693,","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":2021,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.927806Z"},"links":{"cited_paper":"/paper/2310.03693","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:17389c364d032e9de0c27a2fab461f90dd56ee8b695c8fa60e05f35082071d9b","observation_id":"cc261b1b-ca0f-4c9f-81ab-c6d2fb0c1bf6","resolution":{"observed_at":"2026-08-16T06:00:23.927806Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2407.21783","last_updated":"2024-11-23T23:27:33Z","snapshot_observed_at":"2026-08-13T17:20:44.002518Z","submitted_at":"2024-07-31T17:54:27Z","title":"The Llama 3 Herd of Models","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2407.21783","snapshot_observed_at":"2026-08-16T06:00:23.873419Z","title":"The llama 3 herd of models","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":2022,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.873419Z"},"links":{"cited_paper":"/paper/2407.21783","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:ae78a9a65a077f7748a104fcb739c8a2523ed5eb15a6019f111f26b09e61723a","observation_id":"c374fa97-4495-4e8c-8d0c-662716e230c4","resolution":{"observed_at":"2026-08-16T06:00:23.873419Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2403.13187","last_updated":"2025-01-27T10:19:44Z","snapshot_observed_at":"2026-08-16T14:08:13.299258Z","submitted_at":"2024-03-19T22:56:53Z","title":"Evolutionary Optimization of Model Merging Recipes","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2403.13187","snapshot_observed_at":"2026-08-16T06:00:23.856828Z","title":"Takuya Akiba, Makoto Shing, Yujin Tang, Qi Sun, and David Ha","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":2023,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.856828Z"},"links":{"cited_paper":"/paper/2403.13187","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:c24f0bcd09e9153d1c8e73d889d763817ba3859298ee3527bb351ba2bd22e96d","observation_id":"816e993b-776b-427f-9771-d4fff60e4fa9","resolution":{"observed_at":"2026-08-16T06:00:23.856828Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2404.05993","last_updated":"2024-09-11T14:42:29Z","snapshot_observed_at":"2026-08-17T18:25:58.914494Z","submitted_at":"2024-04-09T03:54:28Z","title":"AEGIS: Online Adaptive AI Content Safety Moderation with Ensemble of LLM Experts","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2404.05993","snapshot_observed_at":"2026-08-16T06:00:23.878320Z","title":"Aegis: Online adaptive ai content safety moderation with ensemble of llm experts","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing","version":2},"reference_index":2024,"source":"pdf_text","source_observed_at":"2026-08-16T06:00:23.878320Z"},"links":{"cited_paper":"/paper/2404.05993","citing_paper":"/paper/2504.19333"},"observation_digest":"sha256:7b6d7f38c9b817c2e354153109362cc224ff781c4deb909d7e4e5b932f7ac07c","observation_id":"76a37a7c-027f-4957-9b31-ce521c91672a","resolution":{"observed_at":"2026-08-16T06:00:23.878320Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}}],"paper":{"arxiv_id":"2504.19333","last_updated":"2025-04-29T02:42:00Z","latest_version":2,"primary_category":"cs.CL","snapshot_observed_at":"2026-08-19T01:19:15.165727Z","submitted_at":"2025-04-27T19:07:58Z","title":"Unified Multi-Task Learning & Model Fusion for Efficient Language Model Guardrailing"},"reference_resolution":{"displayed":20,"state_counts":{"malformed_identifier":0,"metadata_mismatch":0,"parse_uncertain":0,"unresolved":18,"verified_exact":0,"verified_fuzzy":2},"total_outbound_references":20},"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":"2026-08-18T06:34:40.430872+00:00","source":"crossref"},{"observed_at":"2026-08-18T06:34:34.496301+00:00","source":"retraction_watch"}],"thesis":"As of 19 August 2026, this Paper Citation Record lists 20 of 20 outbound references and 0 inbound Pith citation observations for arXiv:2504.19333."}