{"as_of":"2026-08-08T06:09:00Z","caps":{"database_statements":6,"inbound":100,"outbound":100},"context_digest":"sha256:eb41011623c5f1869bab7ea5d685cf96ee73b4c2ffe71ff06bda1482afa5e71d","coverage":[{"denominator":28,"lane":"reference_resolution","note":"Typed states for the displayed outbound observations.","records_observed":28,"source":"paper_references, paper_reference_links","source_observed_at":"2026-08-07T10:40:02.073583Z","state":"measured"},{"denominator":32,"lane":"standing_notices","note":"One-hop event checks from named stored sources.","records_observed":32,"source":"scholarly_work_events, retraction_status_cache","source_observed_at":"2026-08-07T06:34:17.273281+00:00","state":"measured"},{"denominator":4,"lane":"inbound_itemization","note":"Pith citing papers itemized under the disclosed page cap.","records_observed":4,"source":"paper_references, paper_reference_links","source_observed_at":"2026-08-04T04:34:33.183294Z","state":"measured"},{"denominator":1,"lane":"external_citation_measurements","note":"A source-named dated measurement, never combined with another source.","records_observed":0,"source":"arxiv_reference","source_observed_at":"2026-06-29T18:43:50.549550Z","state":"measured"}],"external_citation_measurements":[],"inbound":[{"citation":{"cited_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"cited_work":{"arxiv_id":"2506.06384","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2506.06384","snapshot_observed_at":"2026-06-29T18:43:50.549550Z","title":"Detectionmethodforpromptinjectionby integratingpre-trainedmodelandheuristicfeatureengineering","venue":null,"work_id":"bafb7caf-4dad-43fe-a280-1f49af629e67","year":2025},"citing_paper":{"arxiv_id":"2604.27202","last_updated":"2026-04-29T21:09:21Z","snapshot_observed_at":"2026-07-31T11:39:59.649988Z","submitted_at":"2026-04-29T21:09:21Z","title":"Indirect Prompt Injection in the Wild: An Empirical Study of Prevalence, Techniques, and Objectives","version":1},"reference_index":94,"source":"pdf_text","source_observed_at":"2026-05-07T08:55:39.589773Z"},"links":{"cited_paper":"/paper/2506.06384","citing_paper":"/paper/2604.27202"},"observation_digest":"sha256:0243c3ac8d65a145d36369f150b5e70ff1760ec1c97d4ce0dc109d7f5b8852f1","observation_id":"2f6a1569-d3d0-47f7-b50b-a4eacfd6957d","resolution":{"observed_at":"2026-05-12T09:51:29.356507Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-07T06:34:17.273281+00:00","source":"crossref"},{"observed_at":"2026-08-07T06:34:11.927384+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"cited_work":{"arxiv_id":"2506.06384","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2506.06384","snapshot_observed_at":"2026-06-29T18:43:50.549550Z","title":"Detectionmethodforpromptinjectionby integratingpre-trainedmodelandheuristicfeatureengineering","venue":null,"work_id":"bafb7caf-4dad-43fe-a280-1f49af629e67","year":2025},"citing_paper":{"arxiv_id":"2605.26999","last_updated":"2026-05-26T13:19:25Z","snapshot_observed_at":"2026-08-02T09:42:05.440868Z","submitted_at":"2026-05-26T13:19:25Z","title":"Prompt Injection Detection is Regime-Dependent: A Deployment-Aware Evaluation with Interpretable Structural Signals","version":1},"reference_index":14,"source":"pdf_text","source_observed_at":"2026-06-29T18:41:03.566306Z"},"links":{"cited_paper":"/paper/2506.06384","citing_paper":"/paper/2605.26999"},"observation_digest":"sha256:9ea1cec7bd3684e08ac3a07e0b7dcf5db6702b5a83467057fcd7197af19bd108","observation_id":"0792b659-996d-4496-8912-a8fd2971fa16","resolution":{"observed_at":"2026-06-29T18:43:50.550887Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-07T06:34:17.273281+00:00","source":"crossref"},{"observed_at":"2026-08-07T06:34:11.927384+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2506.06384","snapshot_observed_at":"2026-08-02T08:55:49.093096Z","title":"https://arxiv.org/abs/2506.06384","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2607.19396","last_updated":"2026-08-02T14:07:20Z","snapshot_observed_at":"2026-08-06T23:24:50.017918Z","submitted_at":"2026-07-03T14:24:26Z","title":"CrackedPDFs: A Controlled Benchmark for Hidden Prompt Injection in PDFs","version":1},"reference_index":17,"source":"pdf_text","source_observed_at":"2026-08-02T08:55:49.093096Z"},"links":{"cited_paper":"/paper/2506.06384","citing_paper":"/paper/2607.19396"},"observation_digest":"sha256:c0b09b25c9ae7e9b505904ffb8968f544f4d479c6bb72648622d090de8d8c258","observation_id":"f6fdcd99-579b-4d33-9c1c-929a97a216a0","resolution":{"observed_at":"2026-08-02T08:55:49.093096Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2506.06384","snapshot_observed_at":"2026-08-04T04:34:33.183294Z","title":"https://arxiv.org/abs/2506.06384","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2607.19396","last_updated":"2026-08-02T14:07:20Z","snapshot_observed_at":"2026-08-06T23:24:50.017918Z","submitted_at":"2026-07-03T14:24:26Z","title":"CrackedPDFs: A Controlled Benchmark for Hidden Prompt Injection in PDFs","version":2},"reference_index":17,"source":"pdf_text","source_observed_at":"2026-08-04T04:34:33.183294Z"},"links":{"cited_paper":"/paper/2506.06384","citing_paper":"/paper/2607.19396"},"observation_digest":"sha256:face7c41e10a0c8ed6f2d9df95b3a96d88a0843ccf9b5fb61005607b9b1888d9","observation_id":"b4d5d6f6-8c08-4815-9c41-571f3634349c","resolution":{"observed_at":"2026-08-04T04:34:33.183294Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}}],"links":{"evidence":"/evidence","html":"/paper/2506.06384/citation-record","integrity":"/paper/2506.06384/integrity","json":"/paper/2506.06384/citation-record.json","paper":"/paper/2506.06384"},"outbound":[{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T10:40:01.947408Z","title":"Train- ing language models to follow instructions with human feedback.Advances in neural information processing systems, 35:27730–27744, 2022","venue":null,"work_id":null,"year":2022},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":1,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:01.947408Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:85ef6331e06628439fd266fda86e1095c6252350b4ccd7ffb8262069e6746a5e","observation_id":"c4ba760a-c8d5-436b-a4cb-2ce710cb6e8c","resolution":{"observed_at":"2026-08-07T10:40:01.947408Z","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":"2026-08-07T10:40:01.952724Z","title":"Palm: Scaling language modeling with pathways.Journal of Machine Learning Research, 24(240):1–113, 2023","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":2,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:01.952724Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:d9755b828dc8750d8b3d689e8adac8330467589909945c59696b06dc763ff02b","observation_id":"c117b332-c730-4fcd-9a42-205d554d5a50","resolution":{"observed_at":"2026-08-07T10:40:01.952724Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2102.02503","last_updated":"2021-02-04T09:27:04Z","snapshot_observed_at":"2026-08-05T09:16:06.978059Z","submitted_at":"2021-02-04T09:27:04Z","title":"Understanding the Capabilities, Limitations, and Societal Impact of Large Language Models","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2102.02503","snapshot_observed_at":"2026-08-07T10:40:01.957240Z","title":"Understanding the capabilities, limitations, and societal impact of large language models (2021).arXiv preprint arXiv:2102.02503, 2021","venue":null,"work_id":null,"year":2021},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":3,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:01.957240Z"},"links":{"cited_paper":"/paper/2102.02503","citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:0552777072dbbf9684325f2dc3d6ea9067309b7ae8688ad1bff36ac69baaa3e6","observation_id":"d3a47ffe-2f99-498b-9b08-4b510b03df8c","resolution":{"observed_at":"2026-08-07T10:40:01.957240Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2301.11325","last_updated":"2023-01-26T18:58:53Z","snapshot_observed_at":"2026-07-06T14:45:00.730733Z","submitted_at":"2023-01-26T18:58:53Z","title":"MusicLM: Generating Music From Text","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2301.11325","snapshot_observed_at":"2026-08-07T10:40:01.962885Z","title":"Musiclm: Generating music from text.arXiv preprint arXiv:2301.11325, 2023","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":4,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:01.962885Z"},"links":{"cited_paper":"/paper/2301.11325","citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:4e43f53f506cfc7b36aec474c0e9c561797e339adccabdfd800027f615671d7e","observation_id":"4664779d-693b-4677-bab2-09e320795829","resolution":{"observed_at":"2026-08-07T10:40:01.962885Z","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":"2026-08-07T10:40:02.483998Z","title":"Audiogpt: Under- standing and generating speech, music, sound, and talking head","venue":null,"work_id":"f603810c-4c5a-43c9-9a51-f1e1f499ee64","year":2024},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":5,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:01.968120Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:3dd6b1ed54c6956ad3671023cda3e08868ac172bb3c166fa403c8592ac433a5c","observation_id":"8ea882a2-982e-4262-b2a4-4fab70870866","resolution":{"observed_at":"2026-08-07T10:40:02.488225Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-07T06:34:17.273281+00:00","source":"crossref"},{"observed_at":"2026-08-07T06:34:11.927384+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T10:40:02.469852Z","title":"Understanding large-language model (llm)-powered human-robot interaction","venue":null,"work_id":"25b7eec9-fecd-483b-a017-855a9e9cb733","year":2024},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":6,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:01.972582Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:04b34e88f595c652d2288efa382615b696d76c8072bc40492fc5f3c57cf93397","observation_id":"6d042715-2234-494d-a950-556ea77a2db5","resolution":{"observed_at":"2026-08-07T10:40:02.474369Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-07T06:34:17.273281+00:00","source":"crossref"},{"observed_at":"2026-08-07T06:34:11.927384+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T10:40:02.453464Z","title":"Security and privacy challenges of large language models: A survey.ACM Computing Surveys, 57(6):1– 39, 2025","venue":null,"work_id":"d186c499-5eab-4dd8-9893-f6fb0c94e7c4","year":2025},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":7,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:01.977762Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:2871dcbbf51b98b4bee9cfa93259000ac707baaac4864a34bf7010f98b1d249e","observation_id":"178c40c7-2183-423b-98ad-2a0b4039ebf5","resolution":{"observed_at":"2026-08-07T10:40:02.458542Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-07T06:34:17.273281+00:00","source":"crossref"},{"observed_at":"2026-08-07T06:34:11.927384+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":"2026-08-07T10:40:01.981949Z","title":"A survey on large language model (llm) security and privacy: The good, the bad, and the ugly.High-Confidence Computing, page 100211, 2024","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":8,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:01.981949Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:9b2b2483817bb7b73dccc4b9190a85e5bd1ba23497fedaf4ad65de238414e017","observation_id":"5240f838-8660-41d9-887e-5c30325b5db7","resolution":{"observed_at":"2026-08-07T10:40:01.981949Z","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":"2026-08-07T10:40:02.427959Z","title":"Owasp top 10 list for large language models, 2024","venue":null,"work_id":"47401e00-0dce-4722-9b8f-d161d9c3ee1a","year":2024},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":9,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:01.986740Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:6f6c17b08fdf6164a02bc57c9f7e970380cbfca83767a30841b829595cffdbbd","observation_id":"068b22bf-1c41-4592-8d10-4e81798844b7","resolution":{"observed_at":"2026-08-07T10:40:02.433856Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-07T06:34:17.273281+00:00","source":"crossref"},{"observed_at":"2026-08-07T06:34:11.927384+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T10:40:02.411817Z","title":"Ignore previous prompt: Attack techniques for lan- guage models","venue":null,"work_id":"805b7880-a7a2-4a3f-83b5-47fc55c4ed63","year":2022},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":10,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:01.991325Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:9e951d9aff4555dbd5c1702edf03031422573b24560f6e4939bdf5270c76189c","observation_id":"4b39447d-c473-4624-8fac-011f4335ed64","resolution":{"observed_at":"2026-08-07T10:40:02.416679Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-07T06:34:17.273281+00:00","source":"crossref"},{"observed_at":"2026-08-07T06:34:11.927384+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":"2026-08-07T10:40:01.995703Z","title":"Not what you’ve signed up for: Compromising real-world llm-integrated applications with indirect prompt injection","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":11,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:01.995703Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:95f8ae84686c9cf6eb484deef1f83b2eec50da48977e07379afb89462ac5296a","observation_id":"4ac2a4fc-ee9f-4d76-9172-1b61d398450b","resolution":{"observed_at":"2026-08-07T10:40:01.995703Z","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":"2026-08-07T10:40:02.000141Z","title":"Formalizing and benchmarking prompt injection attacks and defenses","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":12,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:02.000141Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:f30a59b763ee91fddbe29eefb5cd4c3d7784b4ea98723b2ee295ac6aa881a548","observation_id":"714dd145-fbfc-4f49-826f-39b2b5836cc7","resolution":{"observed_at":"2026-08-07T10:40:02.000141Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2407.03391","last_updated":"2024-07-03T14:52:09Z","snapshot_observed_at":"2026-07-06T18:41:09.959699Z","submitted_at":"2024-07-03T14:52:09Z","title":"Soft Begging: Modular and Efficient Shielding of LLMs against Prompt Injection and Jailbreaking based on Prompt Tuning","version":1},"cited_work":{"arxiv_id":"2407.03391","doi":null,"metadata_source":"pith","pith_arxiv_id":"2407.03391","snapshot_observed_at":"2026-08-07T10:40:02.188141Z","title":"Soft Begging: Modular and Efficient Shielding of LLMs against Prompt Injection and Jailbreaking based on Prompt Tuning","venue":"cs.CR","work_id":"6bd5d79f-ccca-49a6-8f73-9f59e999ab1c","year":2024},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":13,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:02.004888Z"},"links":{"cited_paper":"/paper/2407.03391","citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:16807b28adf6a96342d82412d68ae0049ee41628bbe6ff318982a314d050e7d1","observation_id":"86d29bf5-8343-41e5-b7b6-6b65800b7715","resolution":{"observed_at":"2026-08-07T10:40:02.196160Z","resolver_source":"local_arxiv","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-07T06:34:17.273281+00:00","source":"crossref"},{"observed_at":"2026-08-07T06:34:11.927384+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T10:40:02.376605Z","title":"Many-shot jail- breaking.Advances in Neural Information Processing Systems, 37:129696–129742, 2025","venue":null,"work_id":"f51eaf68-208d-4f8d-85bb-769b517e3f56","year":2025},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":14,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:02.009451Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:34770a64e9a3fa3f7f03a47b5fea53731672c2da40a92d461e0e82f3f874c198","observation_id":"c35267ab-5f2f-440d-bad6-77b7dcd9db51","resolution":{"observed_at":"2026-08-07T10:40:02.382294Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-07T06:34:17.273281+00:00","source":"crossref"},{"observed_at":"2026-08-07T06:34:11.927384+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":"2026-08-07T10:40:02.014098Z","title":"Deep reinforcement learning from human preferences.Advances in neural information processing systems, 30, 2017","venue":null,"work_id":null,"year":2017},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":15,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:02.014098Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:0a3ed7650040c1c03d5fcb0459cf90f4c8e87b2d8d01598ab6b7e37275dcd6c2","observation_id":"5c5cbb62-22d0-4aba-8e65-fea1c7bffd8c","resolution":{"observed_at":"2026-08-07T10:40:02.014098Z","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":"2026-08-07T10:40:02.019307Z","title":"Direct preference optimization: Your language model is secretly a reward model.Advances in Neural Information Processing Systems, 36:53728–53741, 2023","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":16,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:02.019307Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:05c59646857a61e0149d275116fbfc23551a464fa8ead25d602ce2da06727a6d","observation_id":"1da5feb4-3636-48f3-875d-a58907861f90","resolution":{"observed_at":"2026-08-07T10:40:02.019307Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2303.08774","last_updated":"2024-03-04T06:01:33Z","snapshot_observed_at":"2026-08-07T07:30:12.213965Z","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":"2026-08-07T10:40:02.023821Z","title":"Gpt-4 technical report.arXiv preprint arXiv:2303.08774, 2023","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":17,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:02.023821Z"},"links":{"cited_paper":"/paper/2303.08774","citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:04115f2ca103678ff82bb44c76f148727704989421f1e654270eaa20a94011f8","observation_id":"15018687-e03a-4e0c-87d9-376046691d05","resolution":{"observed_at":"2026-08-07T10:40:02.023821Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2111.09543","last_updated":"2023-03-24T09:17:17Z","snapshot_observed_at":"2026-07-06T02:11:23.670680Z","submitted_at":"2021-11-18T06:48:00Z","title":"DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding Sharing","version":4},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2111.09543","snapshot_observed_at":"2026-08-07T10:40:02.028913Z","title":"Debertav3: Improving deberta us- ing electra-style pre-training with gradient-disentangled embedding sharing.arXiv preprint arXiv:2111.09543, 2021","venue":null,"work_id":null,"year":2021},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":18,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:02.028913Z"},"links":{"cited_paper":"/paper/2111.09543","citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:e6e776708f0d192db4657a6c33d0caf7f410715c3afcd79ce6ecb72a3f3ee52c","observation_id":"be004a1b-f12e-4353-8d4c-f677a2349ff8","resolution":{"observed_at":"2026-08-07T10:40:02.028913Z","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":"2026-08-07T10:40:02.341554Z","title":"fmops/distilbert-prompt-injection, 2024","venue":null,"work_id":"af4e381b-f45e-4b95-a5e6-f7e62a3b60ca","year":2024},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":19,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:02.033575Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:2640849ae672643d9cdb4b6eac8067effd658eb0bb922f59da88a574d48df1ec","observation_id":"8a797b61-fe85-4db6-ba0a-7a458322ca1b","resolution":{"observed_at":"2026-08-07T10:40:02.346150Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-07T06:34:17.273281+00:00","source":"crossref"},{"observed_at":"2026-08-07T06:34:11.927384+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T10:40:02.324503Z","title":"Fine-tuned deberta-v3-base for prompt injection detection, 2024","venue":null,"work_id":"e7820563-f13e-4d8a-8d65-496ffebd15f3","year":2024},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":20,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:02.038110Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:f40a6c01fbf3497e629bbb7097a1c04eac43c70dc39f5b91f45f809b43f19972","observation_id":"d6e28dd2-6c8c-45ec-9235-8c671ff01b87","resolution":{"observed_at":"2026-08-07T10:40:02.330233Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-07T06:34:17.273281+00:00","source":"crossref"},{"observed_at":"2026-08-07T06:34:11.927384+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T10:40:02.307003Z","title":"Safeguard: A benchmark suite for evaluating attacks and defenses on llm safety, 2023","venue":null,"work_id":"3239113d-e2ef-4aa7-ba5a-51458ee7cbe3","year":2023},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":21,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:02.043048Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:2e38829c03b2758fc55a305de2003a08d1872ff72e69b715e72e5c0ef952e5c3","observation_id":"c8af2019-9508-4de7-8345-8407d8fbd83e","resolution":{"observed_at":"2026-08-07T10:40:02.311868Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-07T06:34:17.273281+00:00","source":"crossref"},{"observed_at":"2026-08-07T06:34:11.927384+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2410.22770","last_updated":"2025-03-30T16:39:15Z","snapshot_observed_at":"2026-07-06T19:42:03.739016Z","submitted_at":"2024-10-30T07:39:42Z","title":"InjecGuard: Benchmarking and Mitigating Over-defense in Prompt Injection Guardrail Models","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.22770","snapshot_observed_at":"2026-08-07T10:40:02.046929Z","title":"Injecguard: Benchmarking and mitigating over-defense in prompt injection guardrail models.arXiv preprint arXiv:2410.22770, 2024","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":22,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:02.046929Z"},"links":{"cited_paper":"/paper/2410.22770","citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:8397bf07abfc875a42920c0d9406ed2f802bbb2c287b090794970ebb3d620dde","observation_id":"347b9855-fd2b-4166-9014-f50b5208fba5","resolution":{"observed_at":"2026-08-07T10:40:02.046929Z","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":"2026-08-07T10:40:02.290756Z","title":"Apply- ing pre-trained multilingual bert in embeddings for improved malicious prompt injection attacks detection","venue":null,"work_id":"2a4d6756-ae8e-48d1-8c0d-aa2446e2089c","year":2024},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":23,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:02.051592Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:97ecca8423281e5d1e5bc07a5c25e7aebfa403d1b6b2b7cc50f2ee5f40905309","observation_id":"3f048285-c3ef-412a-99b6-0e3ac8b8e7ea","resolution":{"observed_at":"2026-08-07T10:40:02.295665Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-07T06:34:17.273281+00:00","source":"crossref"},{"observed_at":"2026-08-07T06:34:11.927384+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2402.06363","last_updated":"2024-09-25T19:48:39Z","snapshot_observed_at":"2026-07-06T17:27:58.440355Z","submitted_at":"2024-02-09T12:15:51Z","title":"StruQ: Defending Against Prompt Injection with Structured Queries","version":2},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2402.06363","snapshot_observed_at":"2026-08-07T10:40:02.055581Z","title":"Struq: Defending against prompt injection with structured queries.arXiv preprint arXiv:2402.06363, 2024","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":24,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:02.055581Z"},"links":{"cited_paper":"/paper/2402.06363","citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:faec840270f3e481ceadeb3b630a6a156f47bb0c5f3695ddf01f505aa37be290","observation_id":"69f768fe-609f-4c44-ac91-940c42493a2c","resolution":{"observed_at":"2026-08-07T10:40:02.055581Z","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":"2026-08-07T10:40:02.273785Z","title":"Jatmo: Prompt injection defense by task- specific finetuning","venue":null,"work_id":"776e5fc6-d0ef-469a-a5c2-57fc32f48e55","year":2024},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":25,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:02.059808Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:7217cc3dacde9971be9100be75376d755e28917448a2289bd6cca7eaea1ce00c","observation_id":"0e4b89cd-86d9-4c36-bfd5-8632b3b46801","resolution":{"observed_at":"2026-08-07T10:40:02.279552Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-07T06:34:17.273281+00:00","source":"crossref"},{"observed_at":"2026-08-07T06:34:11.927384+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T10:40:02.259353Z","title":"Llm self defense: By self examina- tion, llms know they are being tricked","venue":null,"work_id":"11e020fb-d62a-4c4b-bc63-93b24ae92446","year":2024},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":26,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:02.064187Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:99cc13cdcfe89141daeca95014dcac03e9174fb228e6f73d45702628dcab606e","observation_id":"ed4ee657-a579-4f77-ab4e-842425669b6d","resolution":{"observed_at":"2026-08-07T10:40:02.263645Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-07T06:34:17.273281+00:00","source":"crossref"},{"observed_at":"2026-08-07T06:34:11.927384+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":"raw_reference","pith_arxiv_id":null,"snapshot_observed_at":"2026-08-07T10:40:02.243532Z","title":"Defending chatgpt against jailbreak attack via self- reminders.Nature Machine Intelligence, 5(12):1486–1496, 2023","venue":null,"work_id":"97810028-934d-41ab-abcf-53bc43440ec4","year":2023},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":27,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:02.068447Z"},"links":{"citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:fe4c14116fdb526477b7999d7a27dab7566a38a91e542914e78de4d9678c64ea","observation_id":"f13cddc9-c7e3-40ba-b56c-d5d270f17f3d","resolution":{"observed_at":"2026-08-07T10:40:02.248336Z","resolver_source":"raw_fallback","status":"verified_fuzzy"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-07T06:34:17.273281+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-07T06:34:17.273281+00:00","source":"crossref"},{"observed_at":"2026-08-07T06:34:11.927384+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2404.13161","last_updated":"2024-04-19T20:11:12Z","snapshot_observed_at":"2026-08-07T13:01:36.699787Z","submitted_at":"2024-04-19T20:11:12Z","title":"CyberSecEval 2: A Wide-Ranging Cybersecurity Evaluation Suite for Large Language Models","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2404.13161","snapshot_observed_at":"2026-08-07T10:40:02.073583Z","title":"I hate human","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering","version":1},"reference_index":28,"source":"pdf_text","source_observed_at":"2026-08-07T10:40:02.073583Z"},"links":{"cited_paper":"/paper/2404.13161","citing_paper":"/paper/2506.06384"},"observation_digest":"sha256:039b4838c4f715350441482d31fac9419b1f18f35f0ea37768abcd19eda9808b","observation_id":"f2d2bbe2-ea9e-4431-9c65-a349c101150b","resolution":{"observed_at":"2026-08-07T10:40:02.073583Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}}],"paper":{"arxiv_id":"2506.06384","last_updated":"2025-06-05T06:01:19Z","latest_version":1,"primary_category":"cs.CL","snapshot_observed_at":"2026-08-07T17:01:39.395444Z","submitted_at":"2025-06-05T06:01:19Z","title":"Detection Method for Prompt Injection by Integrating Pre-trained Model and Heuristic Feature Engineering"},"reference_resolution":{"displayed":28,"state_counts":{"malformed_identifier":0,"metadata_mismatch":0,"parse_uncertain":0,"unresolved":14,"verified_exact":1,"verified_fuzzy":13},"total_outbound_references":28},"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":"2026-08-07T06:34:17.273281+00:00","source":"crossref"},{"observed_at":"2026-08-07T06:34:11.927384+00:00","source":"retraction_watch"}],"thesis":"As of 8 August 2026, this Paper Citation Record lists 28 of 28 outbound references and 4 inbound Pith citation observations for arXiv:2506.06384."}