{"as_of":"2026-08-20T05:18:00Z","caps":{"database_statements":6,"inbound":100,"outbound":100},"context_digest":"sha256:9533f6000c4f92a8fb759260b76277e90d05a8dc3d0e29ddd19c04d376e60d27","coverage":[{"denominator":0,"lane":"reference_resolution","note":"Typed states for the displayed outbound observations.","records_observed":0,"source":"paper_references, paper_reference_links","source_observed_at":null,"state":"measured"},{"denominator":34,"lane":"standing_notices","note":"One-hop event checks from named stored sources.","records_observed":34,"source":"scholarly_work_events, retraction_status_cache","source_observed_at":"2026-08-19T06:32:44.657259+00:00","state":"measured"},{"denominator":34,"lane":"inbound_itemization","note":"Pith citing papers itemized under the disclosed page cap.","records_observed":34,"source":"paper_references, paper_reference_links","source_observed_at":"2026-08-16T04:38:39.301351Z","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":"2026-07-01T23:36:22.936259Z","state":"measured"}],"external_citation_measurements":[],"inbound":[{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-08-11T10:32:28.233502Z","title":"Aligning llms to be robust against prompt injection.arXiv preprint arXiv:2410.05451, 2024","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2412.16555","last_updated":"2025-05-29T14:05:50Z","snapshot_observed_at":"2026-08-18T23:35:37.993619Z","submitted_at":"2024-12-21T09:43:51Z","title":"Divide and Conquer: A Hybrid Strategy Defeats Multimodal Large Language Models","version":3},"reference_index":11,"source":"pdf_text","source_observed_at":"2026-08-11T10:32:28.233502Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2412.16555"},"observation_digest":"sha256:04870601e0abeb4a3e0c67d188ebb029278114d5e4a6a8bc82e1a798305e63ef","observation_id":"6156e6c5-567b-4b99-ba90-e62cfeaae84c","resolution":{"observed_at":"2026-08-11T10:32:28.233502Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-08-10T22:53:20.639962Z","title":"Aligning LLMs to Be Robust Against Prompt Injection","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2501.00517","last_updated":"2024-12-31T16:01:25Z","snapshot_observed_at":"2026-08-17T02:13:31.546943Z","submitted_at":"2024-12-31T16:01:25Z","title":"A Method for Enhancing the Safety of Large Model Generation Based on Multi-dimensional Attack and Defense","version":1},"reference_index":11,"source":"pdf_text","source_observed_at":"2026-08-10T22:53:20.639962Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2501.00517"},"observation_digest":"sha256:9b47749de935fe95924820b801bb36523dafd87a0e41cbd514add44c0734ca8b","observation_id":"f0db6692-c76b-48b3-b8b8-14609638632d","resolution":{"observed_at":"2026-08-10T22:53:20.639962Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-08-08T20:06:58.233116Z","title":"Aligning llms to be robust against prompt injection","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2502.05174","last_updated":"2025-06-10T18:13:09Z","snapshot_observed_at":"2026-08-19T00:43:34.703842Z","submitted_at":"2025-02-07T18:57:49Z","title":"MELON: Provable Defense Against Indirect Prompt Injection Attacks in AI Agents","version":4},"reference_index":5,"source":"arxiv_source","source_observed_at":"2026-08-08T20:06:58.233116Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2502.05174"},"observation_digest":"sha256:c639229d72eec1332802957f229512d2ce6542ecf7b76fbe42173c037ac04ec5","observation_id":"9dae070a-56be-4e5f-b63a-539284747a04","resolution":{"observed_at":"2026-08-08T20:06:58.233116Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":"2410.05451","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-07-01T23:36:22.936259Z","title":"Secalign: Defending against prompt injection with preference optimization","venue":null,"work_id":"ee1040df-3911-42c9-b921-8e69093f4f74","year":2025},"citing_paper":{"arxiv_id":"2502.05206","last_updated":"2026-04-14T16:10:41Z","snapshot_observed_at":"2026-08-18T22:44:13.666151Z","submitted_at":"2025-02-02T05:14:22Z","title":"Safety at Scale: A Comprehensive Survey of Large Model and Agent Safety","version":6},"reference_index":141,"source":"pdf_text","source_observed_at":"2026-05-23T04:39:04.591722Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2502.05206"},"observation_digest":"sha256:2c74fdef96c3fd53d10ca1b1f04beb2af7ae68a5678c089ffb9e48fefb625333","observation_id":"4c818245-27f2-4528-a0c2-76a337aeccac","resolution":{"observed_at":"2026-05-23T04:42:33.779160Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-19T06:32:44.657259+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-19T06:32:44.657259+00:00","source":"crossref"},{"observed_at":"2026-08-19T06:32:39.956319+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":"2410.05451","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-07-01T23:36:22.936259Z","title":"Secalign: Defending against prompt injection with preference optimization","venue":null,"work_id":"ee1040df-3911-42c9-b921-8e69093f4f74","year":2025},"citing_paper":{"arxiv_id":"2504.19793","last_updated":"2025-08-24T03:28:21Z","snapshot_observed_at":"2026-08-14T15:35:38.720321Z","submitted_at":"2025-04-28T13:36:43Z","title":"Prompt Injection Attack to Tool Selection in LLM Agents","version":3},"reference_index":24,"source":"pdf_text","source_observed_at":"2026-05-16T17:08:28.933831Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2504.19793"},"observation_digest":"sha256:7e36043ee43bc015e11f5735fe5f22f1e4623ae694dedab258756c38ec61956e","observation_id":"2a9cecac-da47-40d9-aa64-005a1ff22a92","resolution":{"observed_at":"2026-05-16T17:08:29.020566Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-19T06:32:44.657259+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-19T06:32:44.657259+00:00","source":"crossref"},{"observed_at":"2026-08-19T06:32:39.956319+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":"2410.05451","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-07-01T23:36:22.936259Z","title":"Secalign: Defending against prompt injection with preference optimization","venue":null,"work_id":"ee1040df-3911-42c9-b921-8e69093f4f74","year":2025},"citing_paper":{"arxiv_id":"2504.20472","last_updated":"2026-04-09T06:57:42Z","snapshot_observed_at":"2026-08-13T08:12:27.488850Z","submitted_at":"2025-04-29T07:13:53Z","title":"Robustness via Referencing: Defending against Prompt Injection Attacks by Referencing the Executed Instruction","version":2},"reference_index":7,"source":"pdf_text","source_observed_at":"2026-05-22T19:10:55.009810Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2504.20472"},"observation_digest":"sha256:54eec4878c9a450ca856fa8ec74db7605a75c4e57729d6e772c2bafeeb043a83","observation_id":"454aa68a-fc7b-4753-ad8b-702f8ba5fbfc","resolution":{"observed_at":"2026-05-22T19:11:58.012323Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-19T06:32:44.657259+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-19T06:32:44.657259+00:00","source":"crossref"},{"observed_at":"2026-08-19T06:32:39.956319+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":"2410.05451","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-07-01T23:36:22.936259Z","title":"Secalign: Defending against prompt injection with preference optimization","venue":null,"work_id":"ee1040df-3911-42c9-b921-8e69093f4f74","year":2025},"citing_paper":{"arxiv_id":"2504.20984","last_updated":"2025-09-10T19:03:48Z","snapshot_observed_at":"2026-08-12T23:40:41.340437Z","submitted_at":"2025-04-29T17:55:52Z","title":"ACE: A Security Architecture for LLM-Integrated App Systems","version":3},"reference_index":30,"source":"pdf_text","source_observed_at":"2026-05-22T18:05:50.363944Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2504.20984"},"observation_digest":"sha256:d6ee259f7b239bd676eab2b12085a754d8b506ed983bd5dd4dce14f1f3ebd635","observation_id":"4563f920-b373-4c37-897a-d4a746f7db47","resolution":{"observed_at":"2026-05-22T18:06:54.286509Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-19T06:32:44.657259+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-19T06:32:44.657259+00:00","source":"crossref"},{"observed_at":"2026-08-19T06:32:39.956319+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-08-16T04:38:39.301351Z","title":"Yulin Chen, Haoran Li, Zihao Zheng, Yangqiu Song, Dekai Wu, and Bryan Hooi","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2505.00843","last_updated":"2025-05-01T20:09:48Z","snapshot_observed_at":"2026-08-19T01:54:04.758285Z","submitted_at":"2025-05-01T20:09:48Z","title":"OET: Optimization-based prompt injection Evaluation Toolkit","version":1},"reference_index":5,"source":"pdf_text","source_observed_at":"2026-08-16T04:38:39.301351Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2505.00843"},"observation_digest":"sha256:090d4851494f862d983ec81a8303d7b00290a0375106d7ad87e044a7cfed372f","observation_id":"e7c1170b-fa30-47c5-8ca8-36d4fe4c9503","resolution":{"observed_at":"2026-08-16T04:38:39.301351Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-08-15T23:23:38.685456Z","title":"”SecAlign: Defending Against Prompt Injection with Preference Optimization.” arXiv:2410.05451","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2505.04806","last_updated":"2025-05-13T05:36:34Z","snapshot_observed_at":"2026-08-16T21:53:39.240862Z","submitted_at":"2025-05-07T21:15:40Z","title":"Red Teaming the Mind of the Machine: A Systematic Evaluation of Prompt Injection and Jailbreak Vulnerabilities in LLMs","version":2},"reference_index":11,"source":"pdf_text","source_observed_at":"2026-08-15T23:23:38.685456Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2505.04806"},"observation_digest":"sha256:11b3d0c4a065d384403b2436ff3839ac141dfa53fd45b0cb191c2513b9b2ee1a","observation_id":"8fd3fd14-a348-4069-957d-a83be6e5f729","resolution":{"observed_at":"2026-08-15T23:23:38.685456Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-08-07T14:36:09.746380Z","title":"Aligning llms to be robust against prompt injection","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2505.18333","last_updated":"2025-05-23T19:39:56Z","snapshot_observed_at":"2026-08-07T14:30:46.734293Z","submitted_at":"2025-05-23T19:39:56Z","title":"A Critical Evaluation of Defenses against Prompt Injection Attacks","version":1},"reference_index":7,"source":"pdf_text","source_observed_at":"2026-08-07T14:36:09.746380Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2505.18333"},"observation_digest":"sha256:861ecc53e31de7036d7dea9ee9a6f8b4171af24afeb67c50676cfee8b891cb0a","observation_id":"92a5095f-584a-4555-8586-88a61d03c25b","resolution":{"observed_at":"2026-08-07T14:36:09.746380Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-08-07T11:59:22.064371Z","title":"Secalign: Defending against prompt injection with preference optimization","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2506.01055","last_updated":"2025-06-01T15:48:06Z","snapshot_observed_at":"2026-08-15T10:52:39.381322Z","submitted_at":"2025-06-01T15:48:06Z","title":"Simple Prompt Injection Attacks Can Leak Personal Data Observed by LLM Agents During Task Execution","version":1},"reference_index":7,"source":"pdf_text","source_observed_at":"2026-08-07T11:59:22.064371Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2506.01055"},"observation_digest":"sha256:a2f6c6e4d5cadad6750256a2c1fd397ddb42d6e28520bae14ef6b8d97d26986c","observation_id":"09fdaad2-4160-442b-b84e-2d0dc3a1298b","resolution":{"observed_at":"2026-08-07T11:59:22.064371Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-08-06T18:32:38.828080Z","title":null,"venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2507.07974","last_updated":"2025-08-25T16:49:10Z","snapshot_observed_at":"2026-08-16T18:37:44.319794Z","submitted_at":"2025-07-10T17:51:05Z","title":"Defending Against Prompt Injection With a Few DefensiveTokens","version":2},"reference_index":4,"source":"pdf_text","source_observed_at":"2026-08-06T18:32:38.828080Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2507.07974"},"observation_digest":"sha256:00a8d0b5900f24d00af253d29dbf09cb322f58c1e1cc036a923392a960855cee","observation_id":"6ede77e6-208c-4571-8ddc-ee3fe5407b0c","resolution":{"observed_at":"2026-08-06T18:32:38.828080Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":"2410.05451","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-07-01T23:36:22.936259Z","title":"Secalign: Defending against prompt injection with preference optimization","venue":null,"work_id":"ee1040df-3911-42c9-b921-8e69093f4f74","year":2025},"citing_paper":{"arxiv_id":"2507.10610","last_updated":"2026-04-07T11:46:55Z","snapshot_observed_at":"2026-08-16T01:05:49.667917Z","submitted_at":"2025-07-13T08:36:09Z","title":"LaSM: Layer-wise Scaling Mechanism for Defending Pop-up Attack on GUI Agents","version":3},"reference_index":12,"source":"pdf_text","source_observed_at":"2026-05-19T04:10:57.882345Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2507.10610"},"observation_digest":"sha256:64af8278317d48ddfa1ee57b2f74571f05873776b2babbb68ab2c427319cd73e","observation_id":"b79bbf34-94e8-487c-bfad-835511aa37cf","resolution":{"observed_at":"2026-05-19T04:12:02.170806Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-19T06:32:44.657259+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-19T06:32:44.657259+00:00","source":"crossref"},{"observed_at":"2026-08-19T06:32:39.956319+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-08-05T10:44:10.399166Z","title":null,"venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.03787","last_updated":"2025-09-04T00:45:58Z","snapshot_observed_at":"2026-08-18T00:30:11.532287Z","submitted_at":"2025-09-04T00:45:58Z","title":"Evaluating the Robustness of Retrieval-Augmented Generation to Adversarial Evidence in the Health Domain","version":1},"reference_index":12,"source":"pdf_text","source_observed_at":"2026-08-05T10:44:10.399166Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2509.03787"},"observation_digest":"sha256:c549ae0cc4cae167f9c64ab52672d9507d4893a51b60c610731bb5820f141c95","observation_id":"16634dc5-cecf-4163-a0d7-286a5723c790","resolution":{"observed_at":"2026-08-05T10:44:10.399166Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-08-04T21:33:57.102424Z","title":null,"venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2509.07941","last_updated":"2025-09-09T17:21:20Z","snapshot_observed_at":"2026-08-19T02:00:41.741397Z","submitted_at":"2025-09-09T17:21:20Z","title":"ImportSnare: Directed \"Code Manual\" Hijacking in Retrieval-Augmented Code Generation","version":1},"reference_index":5,"source":"pdf_text","source_observed_at":"2026-08-04T21:33:57.102424Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2509.07941"},"observation_digest":"sha256:34662ef0d4ede7bae97a1ecb078c7e6f48cbad0a87d905eaacd6628363f6350e","observation_id":"555e7fec-8fa6-4ac4-badd-e5a12015c9d3","resolution":{"observed_at":"2026-08-04T21:33:57.102424Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-08-04T08:06:09.522680Z","title":"Secalign: Defending against prompt injection with preference optimization.arXiv preprint arXiv:2410.05451, 2024","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2510.22963","last_updated":"2026-06-19T20:25:06Z","snapshot_observed_at":"2026-08-14T17:16:52.193930Z","submitted_at":"2025-10-27T03:37:41Z","title":"When Compression Becomes an Attack Surface: Black-Box Attacks on Prompt-Compressed LLM Agents","version":4},"reference_index":23,"source":"pdf_text","source_observed_at":"2026-08-04T08:06:09.522680Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2510.22963"},"observation_digest":"sha256:1cf310ff7bddf27b28cea670566c7fa3ff3f32793d71aec3dd206c949e5fa857","observation_id":"5c262896-4410-4d02-9c69-4315c92d971e","resolution":{"observed_at":"2026-08-04T08:06:09.522680Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":"2410.05451","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-07-01T23:36:22.936259Z","title":"Secalign: Defending against prompt injection with preference optimization","venue":null,"work_id":"ee1040df-3911-42c9-b921-8e69093f4f74","year":2025},"citing_paper":{"arxiv_id":"2510.23883","last_updated":"2026-04-03T16:27:34Z","snapshot_observed_at":"2026-08-13T20:52:15.127327Z","submitted_at":"2025-10-27T21:48:11Z","title":"Agentic AI Security: Threats, Defenses, Evaluation, and Open Challenges","version":3},"reference_index":180,"source":"pdf_text","source_observed_at":"2026-05-18T03:42:10.703369Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2510.23883"},"observation_digest":"sha256:f6dfda44cfb280a1af874e690207c93c3d05b39eaf42419e8c70b66cb1f5dd2e","observation_id":"dcc30560-0954-4523-9158-2be89eac7dd3","resolution":{"observed_at":"2026-05-18T03:42:21.914241Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-19T06:32:44.657259+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-19T06:32:44.657259+00:00","source":"crossref"},{"observed_at":"2026-08-19T06:32:39.956319+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":"2410.05451","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-07-01T23:36:22.936259Z","title":"Secalign: Defending against prompt injection with preference optimization","venue":null,"work_id":"ee1040df-3911-42c9-b921-8e69093f4f74","year":2025},"citing_paper":{"arxiv_id":"2512.00520","last_updated":"2026-04-27T13:00:45Z","snapshot_observed_at":"2026-08-16T18:19:55.894013Z","submitted_at":"2025-11-29T15:31:16Z","title":"Toward a Safe Internet of Agents","version":2},"reference_index":4,"source":"pdf_text","source_observed_at":"2026-05-17T03:16:40.622915Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2512.00520"},"observation_digest":"sha256:733c219220efeb64da17aea04ffefa286549ba843ea35baf62d53abff3bebf8c","observation_id":"35d43ed6-4797-4df0-b6b8-c14e88ab66be","resolution":{"observed_at":"2026-05-17T03:18:56.885802Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-19T06:32:44.657259+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-19T06:32:44.657259+00:00","source":"crossref"},{"observed_at":"2026-08-19T06:32:39.956319+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-07-13T12:52:49.101462Z","title":null,"venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2604.03598","last_updated":"2026-04-04T05:49:30Z","snapshot_observed_at":"2026-08-13T19:43:16.224753Z","submitted_at":"2026-04-04T05:49:30Z","title":"AttackEval: A Systematic Empirical Study of Prompt Injection Attack Effectiveness Against Large Language Models","version":1},"reference_index":6,"source":"pdf_text","source_observed_at":"2026-07-13T12:52:49.101462Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2604.03598"},"observation_digest":"sha256:6ed43bd2d55c6d593a97d20533778bef286c79d85ae3f06cb0e8461c71ba0255","observation_id":"566d8f0b-9332-4c18-8dac-7c8b7a4f5eaa","resolution":{"observed_at":"2026-07-13T12:52:49.101462Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":"2410.05451","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-07-01T23:36:22.936259Z","title":"Secalign: Defending against prompt injection with preference optimization","venue":null,"work_id":"ee1040df-3911-42c9-b921-8e69093f4f74","year":2025},"citing_paper":{"arxiv_id":"2604.12817","last_updated":"2026-04-14T14:43:55Z","snapshot_observed_at":"2026-08-17T09:51:46.671342Z","submitted_at":"2026-04-14T14:43:55Z","title":"Understanding and Improving Continuous Adversarial Training for LLMs via In-context Learning Theory","version":1},"reference_index":3,"source":"pdf_text","source_observed_at":"2026-05-10T15:54:59.122635Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2604.12817"},"observation_digest":"sha256:2d2c71a47238d94ab35eecc7dc59e75172e92dd372e114ed056f82dd05117404","observation_id":"bfa1f9af-83a5-4e2b-a4c6-6d8c334b1dd6","resolution":{"observed_at":"2026-05-11T09:36:10.585625Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-19T06:32:44.657259+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-19T06:32:44.657259+00:00","source":"crossref"},{"observed_at":"2026-08-19T06:32:39.956319+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":"2410.05451","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-07-01T23:36:22.936259Z","title":"Secalign: Defending against prompt injection with preference optimization","venue":null,"work_id":"ee1040df-3911-42c9-b921-8e69093f4f74","year":2025},"citing_paper":{"arxiv_id":"2604.25109","last_updated":"2026-04-28T01:32:27Z","snapshot_observed_at":"2026-08-12T14:49:16.660670Z","submitted_at":"2026-04-28T01:32:27Z","title":"Structured Security Auditing and Robustness Enhancement for Untrusted Agent Skills","version":1},"reference_index":5,"source":"pdf_text","source_observed_at":"2026-05-07T16:22:49.737626Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2604.25109"},"observation_digest":"sha256:139915d6c184d2b6e06c42943ff9c4e4b3f9d478483bf6eb27bdfceecabe4fac","observation_id":"377c23ef-e515-45a9-9e0c-cbc3f560a1f6","resolution":{"observed_at":"2026-05-11T23:46:17.200381Z","resolver_source":"arxiv_id","status":"metadata_mismatch"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-19T06:32:44.657259+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-19T06:32:44.657259+00:00","source":"crossref"},{"observed_at":"2026-08-19T06:32:39.956319+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":"2410.05451","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-07-01T23:36:22.936259Z","title":"Secalign: Defending against prompt injection with preference optimization","venue":null,"work_id":"ee1040df-3911-42c9-b921-8e69093f4f74","year":2025},"citing_paper":{"arxiv_id":"2605.01078","last_updated":"2026-05-01T20:22:40Z","snapshot_observed_at":"2026-08-12T22:41:49.054082Z","submitted_at":"2026-05-01T20:22:40Z","title":"A Sentence Relation-Based Approach to Sanitizing Malicious Instructions","version":1},"reference_index":4,"source":"pdf_text","source_observed_at":"2026-05-09T18:40:16.916424Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2605.01078"},"observation_digest":"sha256:7f3f82c82ec56bc09d2b1b8a815314d10c97bd5aba9e52a32f2071e38f2ceb58","observation_id":"af047c79-62ae-43c9-ba05-b18cbbb12463","resolution":{"observed_at":"2026-05-11T16:06:43.365091Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-19T06:32:44.657259+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-19T06:32:44.657259+00:00","source":"crossref"},{"observed_at":"2026-08-19T06:32:39.956319+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":"2410.05451","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-07-01T23:36:22.936259Z","title":"Secalign: Defending against prompt injection with preference optimization","venue":null,"work_id":"ee1040df-3911-42c9-b921-8e69093f4f74","year":2025},"citing_paper":{"arxiv_id":"2605.11868","last_updated":"2026-05-12T09:48:53Z","snapshot_observed_at":"2026-07-06T23:23:39.723268Z","submitted_at":"2026-05-12T09:48:53Z","title":"IPI-proxy: An Intercepting Proxy for Red-Teaming Web-Browsing AI Agents Against Indirect Prompt Injection","version":1},"reference_index":17,"source":"pdf_text","source_observed_at":"2026-05-13T05:46:07.132408Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2605.11868"},"observation_digest":"sha256:fa7e1751832249e83dd12d8f756b8800eb49b582cabd8b5f6402cf3909e5590a","observation_id":"2d6c8c29-bc32-4478-96af-0c0e02594d39","resolution":{"observed_at":"2026-05-13T05:47:21.337632Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-19T06:32:44.657259+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-19T06:32:44.657259+00:00","source":"crossref"},{"observed_at":"2026-08-19T06:32:39.956319+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":"2410.05451","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-07-01T23:36:22.936259Z","title":"Secalign: Defending against prompt injection with preference optimization","venue":null,"work_id":"ee1040df-3911-42c9-b921-8e69093f4f74","year":2025},"citing_paper":{"arxiv_id":"2605.14290","last_updated":"2026-05-14T02:48:57Z","snapshot_observed_at":"2026-08-19T05:57:11.916502Z","submitted_at":"2026-05-14T02:48:57Z","title":"Web Agents Should Adopt the Plan-Then-Execute Paradigm","version":1},"reference_index":5,"source":"pdf_text","source_observed_at":"2026-05-15T02:42:05.644536Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2605.14290"},"observation_digest":"sha256:c579011df3134ae297542e2c73333215d85ce0611ecb87e2bd17cca91bf7037d","observation_id":"90beb3a0-1e9f-4bf0-bfe7-3894b5c18d73","resolution":{"observed_at":"2026-05-15T02:43:33.387927Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-19T06:32:44.657259+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-19T06:32:44.657259+00:00","source":"crossref"},{"observed_at":"2026-08-19T06:32:39.956319+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":"2410.05451","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-07-01T23:36:22.936259Z","title":"Secalign: Defending against prompt injection with preference optimization","venue":null,"work_id":"ee1040df-3911-42c9-b921-8e69093f4f74","year":2025},"citing_paper":{"arxiv_id":"2606.02668","last_updated":"2026-06-01T11:08:17Z","snapshot_observed_at":"2026-07-06T23:43:02.654629Z","submitted_at":"2026-06-01T11:08:17Z","title":"What You Approve Is What Executes: Consent Integrity for Black-Box LLM Agents","version":1},"reference_index":22,"source":"pdf_text","source_observed_at":"2026-06-28T14:12:38.815852Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2606.02668"},"observation_digest":"sha256:7a12b1489917b1f8154398d5554747d7121dc8c4098d94a92acf2826b5e3b5f3","observation_id":"23b62fb7-766d-4375-b67c-6e2f96052aec","resolution":{"observed_at":"2026-07-01T23:36:22.938900Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-19T06:32:44.657259+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-19T06:32:44.657259+00:00","source":"crossref"},{"observed_at":"2026-08-19T06:32:39.956319+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":"2410.05451","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-07-01T23:36:22.936259Z","title":"Secalign: Defending against prompt injection with preference optimization","venue":null,"work_id":"ee1040df-3911-42c9-b921-8e69093f4f74","year":2025},"citing_paper":{"arxiv_id":"2606.28739","last_updated":"2026-06-27T05:26:43Z","snapshot_observed_at":"2026-08-08T08:45:01.144647Z","submitted_at":"2026-06-27T05:26:43Z","title":"Agent Safety Is Action Alignment","version":1},"reference_index":4,"source":"pdf_text","source_observed_at":"2026-06-30T09:50:45.759936Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2606.28739"},"observation_digest":"sha256:a052539d341091937a29f6fdb63a06de10a6803a6aa1f44a58a7c8d68f342369","observation_id":"b0e91c8f-a793-4727-826e-d19e4a2f315b","resolution":{"observed_at":"2026-06-30T09:54:34.990454Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-19T06:32:44.657259+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-19T06:32:44.657259+00:00","source":"crossref"},{"observed_at":"2026-08-19T06:32:39.956319+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":"2410.05451","doi":null,"metadata_source":"arxiv_reference","pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-07-01T23:36:22.936259Z","title":"Secalign: Defending against prompt injection with preference optimization","venue":null,"work_id":"ee1040df-3911-42c9-b921-8e69093f4f74","year":2025},"citing_paper":{"arxiv_id":"2606.30383","last_updated":"2026-06-29T14:39:38Z","snapshot_observed_at":"2026-08-04T03:27:44.496352Z","submitted_at":"2026-06-29T14:39:38Z","title":"Whose Side Is Your Agent On? Multi-Party Principal Loyalty in LLM Agents","version":1},"reference_index":10,"source":"pdf_text","source_observed_at":"2026-06-30T05:48:15.990443Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2606.30383"},"observation_digest":"sha256:7a9a2b8635344e4c93bf5d2edc3fba085d12dbe570e7429f94751d52f9a16232","observation_id":"99b5583a-a398-4e35-9086-452c7455a7c7","resolution":{"observed_at":"2026-06-30T13:44:41.472646Z","resolver_source":"arxiv_id","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-19T06:32:44.657259+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-19T06:32:44.657259+00:00","source":"crossref"},{"observed_at":"2026-08-19T06:32:39.956319+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-08-02T11:28:37.338330Z","title":"SecAlign: Defending against prompt injection with preference opti- mization, 2024","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2607.20494","last_updated":"2026-06-12T18:54:24Z","snapshot_observed_at":"2026-08-18T17:30:52.370219Z","submitted_at":"2026-06-12T18:54:24Z","title":"Isolating LLM Alignment from Regex: Zero Coverage and Metric-Dependent Divergence Under Adversarial Mutation","version":1},"reference_index":5,"source":"pdf_text","source_observed_at":"2026-08-02T11:28:37.338330Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2607.20494"},"observation_digest":"sha256:168ea530dca8a9c01aeedc655c814c8c68259278f788defd2605fd7b417552c9","observation_id":"17fc53ab-c396-4b37-a67a-11cde6e242d2","resolution":{"observed_at":"2026-08-02T11:28:37.338330Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-07-31T23:24:19.513672Z","title":null,"venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2607.23999","last_updated":"2026-07-28T07:27:46Z","snapshot_observed_at":"2026-08-15T13:44:43.771184Z","submitted_at":"2026-07-27T04:51:20Z","title":"ContainmentBench: Trace-Based Evaluation of Post-Injection Containment in Tool-Using LLM Agents","version":2},"reference_index":6,"source":"pdf_text","source_observed_at":"2026-07-31T23:24:19.513672Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2607.23999"},"observation_digest":"sha256:0f86bb1741c03d3362471d922a7bd2ce81bfaeb586c9528d2ffdbb4ced0c3a5d","observation_id":"7e004625-42f8-46f4-ada6-d0988f2534ca","resolution":{"observed_at":"2026-07-31T23:24:19.513672Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-08-08T00:51:37.929904Z","title":"arXiv preprint arXiv:2410.05451 , year=","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2608.04053","last_updated":"2026-08-04T10:41:34Z","snapshot_observed_at":"2026-08-18T23:16:16.351226Z","submitted_at":"2026-08-04T10:41:34Z","title":"AgentAntibody: An Adaptive Immune System for Defending LLM Agents against Prompt Injection","version":1},"reference_index":25,"source":"arxiv_source","source_observed_at":"2026-08-08T00:51:37.929904Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2608.04053"},"observation_digest":"sha256:5205ad9725a8b981fa4e54e062141715dd0c9adfdf4b106c3d7756d8e059b422","observation_id":"7b9fdc99-be04-4fb2-98ee-3af458b5eefb","resolution":{"observed_at":"2026-08-08T00:51:37.929904Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-08-08T10:29:59.061970Z","title":"arXiv preprint arXiv:2410.05451 , year =","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2608.05563","last_updated":"2026-08-07T09:49:25Z","snapshot_observed_at":"2026-08-18T03:27:47.064066Z","submitted_at":"2026-08-06T03:32:43Z","title":"When Experience Becomes Instruction: Trajectory Poisoning in Self-Evolving Agent Skill Systems","version":1},"reference_index":23,"source":"arxiv_source","source_observed_at":"2026-08-08T10:29:59.061970Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2608.05563"},"observation_digest":"sha256:2f7060000c718ffea2b0e805721e3702ebabbc942943586ba863728d7f5b7ed9","observation_id":"f3623add-4284-47d0-8bfa-52234f36fd50","resolution":{"observed_at":"2026-08-08T10:29:59.061970Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-08-10T04:29:49.865067Z","title":"arXiv preprint arXiv:2410.05451 , year =","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2608.05563","last_updated":"2026-08-07T09:49:25Z","snapshot_observed_at":"2026-08-18T03:27:47.064066Z","submitted_at":"2026-08-06T03:32:43Z","title":"When Experience Becomes Instruction: Trajectory Poisoning in Self-Evolving Agent Skill Systems","version":2},"reference_index":23,"source":"arxiv_source","source_observed_at":"2026-08-10T04:29:49.865067Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2608.05563"},"observation_digest":"sha256:8c8dfc78e7c23b6b1a5698fe4b8527c0d8fb85ddac41828b05691c59c129e99f","observation_id":"669dca24-3b73-4494-9bb7-73974c2fc073","resolution":{"observed_at":"2026-08-10T04:29:49.865067Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-08-15T14:21:41.862200Z","title":null,"venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2608.10530","last_updated":"2026-08-11T06:11:26Z","snapshot_observed_at":"2026-08-19T18:52:42.862550Z","submitted_at":"2026-08-11T06:11:26Z","title":"On Understanding, Identifying, and Mitigating Vulnerabilities in Agentic Large Language Models","version":1},"reference_index":17,"source":"pdf_text","source_observed_at":"2026-08-15T14:21:41.862200Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2608.10530"},"observation_digest":"sha256:59baac1b1fc7aaae5c7468755a2930d962909dc842a90e6097974415f445de3a","observation_id":"c5f3f421-b485-4e17-bab6-46092d2d418e","resolution":{"observed_at":"2026-08-15T14:21:41.862200Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}},{"citation":{"cited_paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization","version":3},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2410.05451","snapshot_observed_at":"2026-08-15T19:23:23.073914Z","title":"Secalign: Defending against prompt injection with preference optimization","venue":null,"work_id":null,"year":2025},"citing_paper":{"arxiv_id":"2608.12977","last_updated":"2026-08-13T08:57:04Z","snapshot_observed_at":"2026-08-18T00:17:46.341587Z","submitted_at":"2026-08-13T08:57:04Z","title":"Beyond Handcrafted Security: Towards Self-Evolving Defense for LLM Agents","version":1},"reference_index":12,"source":"pdf_text","source_observed_at":"2026-08-15T19:23:23.073914Z"},"links":{"cited_paper":"/paper/2410.05451","citing_paper":"/paper/2608.12977"},"observation_digest":"sha256:1b0a2f9bec8418a97655e46c5ced3aa35b0e3d0c54f3901bd66ef9c818ed7d54","observation_id":"a59a9ec2-571e-4052-9299-c54acff204a3","resolution":{"observed_at":"2026-08-15T19:23:23.073914Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}}],"links":{"evidence":"/evidence","html":"/paper/2410.05451/citation-record","integrity":"/paper/2410.05451/integrity","json":"/paper/2410.05451/citation-record.json","paper":"/paper/2410.05451"},"outbound":[],"paper":{"arxiv_id":"2410.05451","last_updated":"2025-07-03T05:45:41Z","latest_version":3,"primary_category":"cs.CR","snapshot_observed_at":"2026-08-16T13:11:45.940997Z","submitted_at":"2024-10-07T19:34:35Z","title":"SecAlign: Defending Against Prompt Injection with Preference Optimization"},"reference_resolution":{"displayed":0,"state_counts":{"malformed_identifier":0,"metadata_mismatch":0,"parse_uncertain":0,"unresolved":0,"verified_exact":0,"verified_fuzzy":0},"total_outbound_references":0},"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":"2026-08-19T06:32:44.657259+00:00","source":"crossref"},{"observed_at":"2026-08-19T06:32:39.956319+00:00","source":"retraction_watch"}],"thesis":"As of 20 August 2026, this Paper Citation Record lists 0 of 0 outbound references and 34 inbound Pith citation observations for arXiv:2410.05451."}