{"as_of":"2026-08-21T03:26:00Z","caps":{"database_statements":6,"inbound":100,"outbound":100},"context_digest":"sha256:93cfab78114f4ea7460264f1eedab56eb6b9f46d8fc1f377d3329f0c7e3e53f4","coverage":[{"denominator":18,"lane":"reference_resolution","note":"Typed states for the displayed outbound observations.","records_observed":18,"source":"paper_references, paper_reference_links","source_observed_at":"2026-06-29T21:16:48.677159Z","state":"measured"},{"denominator":20,"lane":"standing_notices","note":"One-hop event checks from named stored sources.","records_observed":20,"source":"scholarly_work_events, retraction_status_cache","source_observed_at":"2026-08-20T06:33:59.587034+00:00","state":"measured"},{"denominator":2,"lane":"inbound_itemization","note":"Pith citing papers itemized under the disclosed page cap.","records_observed":2,"source":"paper_references, paper_reference_links","source_observed_at":"2026-08-15T22:04:21.540541Z","state":"measured"},{"denominator":1,"lane":"external_citation_measurements","note":"A source-named dated measurement, never combined with another source.","records_observed":0,"source":"pith","source_observed_at":"2026-07-04T14:09:53.283015Z","state":"measured"}],"external_citation_measurements":[],"inbound":[{"citation":{"cited_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"cited_work":{"arxiv_id":"2605.26269","doi":null,"metadata_source":"pith","pith_arxiv_id":"2605.26269","snapshot_observed_at":"2026-07-04T14:09:53.283015Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","venue":"cs.CR","work_id":"e104a193-765b-4642-9604-428ca68636ca","year":2026},"citing_paper":{"arxiv_id":"2606.26627","last_updated":"2026-06-25T05:44:18Z","snapshot_observed_at":"2026-08-14T08:17:57.360957Z","submitted_at":"2026-06-25T05:44:18Z","title":"Agents That Know Too Much: A Data-Centric Survey of Privacy in LLM Agents","version":1},"reference_index":5,"source":"pdf_text","source_observed_at":"2026-06-26T04:29:16.386339Z"},"links":{"cited_paper":"/paper/2605.26269","citing_paper":"/paper/2606.26627"},"observation_digest":"sha256:7a3c146e686d50462e9a7479b0760681b161b850ce84da77cd9abc5f5687c17c","observation_id":"7fa13487-93ae-4a6e-a7c4-08c76100a160","resolution":{"observed_at":"2026-07-04T14:09:53.284810Z","resolver_source":"local_arxiv","status":"verified_exact"},"standing_notice":{"events":[],"observation":"No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.","reason":null,"source_receipts":[{"observed_at":"2026-08-20T06:33:59.587034+00:00","source":"crossref"},{"observed_at":"2026-08-20T06:33:54.927442+00:00","source":"retraction_watch"}],"state":"measured"}},{"citation":{"cited_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":"2605.26269","snapshot_observed_at":"2026-08-15T22:04:21.540541Z","title":"arXiv preprint arXiv:2605.26269 , year=","venue":null,"work_id":null,"year":null},"citing_paper":{"arxiv_id":"2608.12851","last_updated":"2026-08-13T05:47:43Z","snapshot_observed_at":"2026-08-18T06:39:51.703194Z","submitted_at":"2026-08-13T05:47:43Z","title":"Practice Makes Unsafe: Skill Misevolution in Self-Improving LLM Agents","version":1},"reference_index":38,"source":"arxiv_source","source_observed_at":"2026-08-15T22:04:21.540541Z"},"links":{"cited_paper":"/paper/2605.26269","citing_paper":"/paper/2608.12851"},"observation_digest":"sha256:eb92685448f1c8fd03daf5702286fd1cc6c92ba70faced95e8ff746e988042a7","observation_id":"9be6a542-8daa-471e-8f38-2277f3d7ba5f","resolution":{"observed_at":"2026-08-15T22:04:21.540541Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}}],"links":{"evidence":"/evidence","html":"/paper/2605.26269/citation-record","integrity":"/paper/2605.26269/integrity","json":"/paper/2605.26269/citation-record.json","paper":"/paper/2605.26269"},"outbound":[{"citation":{"cited_paper":null,"cited_work":{"arxiv_id":null,"doi":null,"metadata_source":null,"pith_arxiv_id":null,"snapshot_observed_at":"2026-06-29T21:16:48.677159Z","title":"Tensor Trust: Interpretable prompt injection attacks from an online game,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":1,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:25227d2fced12bcd64a5812a349df3a70ca68532a09b8515ca6f3918c06698a9","observation_id":"72b62889-3a51-4557-a736-87959e31ac3c","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","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":"2026-06-29T21:16:48.677159Z","title":"AgentDojo: A dynamic environment to evaluate prompt injection at- tacks and defenses for LLM agents,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":2,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:b09f9e06d8c40967c7883227bc69d0c9b15d3e5570e58a46967a00ee942f8569","observation_id":"f1d684b3-ff6f-45da-9fa1-3efc333aa256","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","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":"2026-06-29T21:16:48.677159Z","title":"Retrieval-augmented generation for knowledge-intensive NLP tasks,","venue":null,"work_id":null,"year":2020},"citing_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":3,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:aaf7f7d6b32672aaa5cc15ba2c3b6c672425eab50599ce2c775358cccb026d64","observation_id":"3975ac22-6c29-4888-a344-8290da4fc525","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","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":"2026-06-29T21:16:48.677159Z","title":"Dense passage retrieval for open-domain question answering,","venue":null,"work_id":null,"year":2020},"citing_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":4,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:0ccf47cf74ce326fda80aa03909e894484b12283717a04b8b4f070250a983e0c","observation_id":"3fa95d3e-3aa2-47a8-a837-7e5eb2826c52","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","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":"2026-06-29T21:16:48.677159Z","title":"Extracting training data from large language models,","venue":null,"work_id":null,"year":2021},"citing_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":5,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:917e6edc3a48b06d116f5fed1a675bd569beb4c3898e0c4ccffee1d8d3ce5dd3","observation_id":"ba7c5521-56d3-4223-b1e1-fdef7ca3e7bd","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","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":"2026-06-29T21:16:48.677159Z","title":"ReAct: Synergizing reasoning and acting in language models,","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":6,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:7f258c61e9e5bc8c4e2d4037bf2df9fe0252bb51a3f1671f31eb5fb05bb015b1","observation_id":"a7f658cd-2c68-4b6d-a6e9-389750daa030","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","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":"2026-06-29T21:16:48.677159Z","title":"Toolformer: Language models can teach themselves to use tools,","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":7,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:405fc57a87726d259d68b17914a3ed6806883400b8f46018c0d0a7652539345a","observation_id":"e57f61ee-7138-477f-b2c5-7621c50bba50","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","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":"2026-06-29T21:16:48.677159Z","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":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":8,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:564d59279278e908533b2cdaaa16d7eb90325311a4656514fe2381285181a784","observation_id":"7e3b1c49-756f-4c7f-bc1d-7f3763901688","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","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":"2026-06-29T21:16:48.677159Z","title":"Formalizing and benchmarking prompt injection attacks and defenses,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":9,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:52ad37b818dee7f488fcbfcd63669e6b98df7107d5669469ba01140a2b3d7d75","observation_id":"4730b1fe-5a19-43f9-b2a5-42b3ac01d75e","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","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":"2026-06-29T21:16:48.677159Z","title":"InjecAgent: Benchmarking indirect prompt injections in tool-integrated large language model agents,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":10,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:e5c9535abe838dedbbb125c779a3abccf8305d5426f25513b33cc44fe963048a","observation_id":"601e5bf5-1fa1-4fbc-ba71-c35c6d366634","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","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":"2026-06-29T21:16:48.677159Z","title":"Security policies and security models,","venue":null,"work_id":null,"year":1982},"citing_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":11,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:b8d3e92e1bd672e55518d4860b818a5cb24f7faf6214b4bfffd588970db81900","observation_id":"8253c0bd-53d1-45c6-9e61-58270ede9c5f","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","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":"2026-06-29T21:16:48.677159Z","title":"Language-based information-flow security,","venue":null,"work_id":null,"year":2003},"citing_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":12,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:e83a3c6c6024a664c3d4105c009a167741d3ef37c2d4b964bf9d476b19f0b7b9","observation_id":"35a6cd26-9e68-4dab-aace-1f410191ba1c","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","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":"2026-06-29T21:16:48.677159Z","title":"Universally composable security: A new paradigm for cryptographic protocols,","venue":null,"work_id":null,"year":2001},"citing_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":13,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:9674c98a4ad274574181269cc417b6a3a10e83e2aa472a0043bc0665707e76ce","observation_id":"e1b540df-fe38-471e-9500-b732d30d7cf7","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","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":"2026-06-29T21:16:48.677159Z","title":"Universal adversarial triggers for attacking and analyzing NLP,","venue":null,"work_id":null,"year":2019},"citing_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":14,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:441c0bfe34bb439b7bb09554bd397d5088fe0297c3f5faa36c40234d91207143","observation_id":"33bc4937-7f92-45ae-ab9c-0c8904a95629","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","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":"2026-06-29T21:16:48.677159Z","title":"Universal and transferable ad- versarial attacks on aligned language models,","venue":null,"work_id":null,"year":2024},"citing_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":15,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:c62c97756f03686038bfc651a8c61d8460935c117194b1c8fbf94cb9aaee2023","observation_id":"e9bb1629-2aca-4b79-b6a3-c86ce8e9c520","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","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":"2026-06-29T21:16:48.677159Z","title":"Quantifying memorization across neural language models,","venue":null,"work_id":null,"year":2023},"citing_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":16,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:7e8273ca830a14eee034990853613fde80f6b23d38ad6a99a6147e6e05192d39","observation_id":"f17b21ec-8260-49d2-9b9f-ea7eff8a7ad7","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","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":"2026-06-29T21:16:48.677159Z","title":"Deduplicating training data mitigates privacy risks in language models,","venue":null,"work_id":null,"year":2022},"citing_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":17,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:9353d43ce54bbd9551f9532446e3d62ac751fa2414f5c3f8993c699499b4e5fd","observation_id":"81fe9754-35d7-467b-bddc-b31e57a79621","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","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":"2026-06-29T21:16:48.677159Z","title":"Training language models to follow instructions with human feedback,","venue":null,"work_id":null,"year":2022},"citing_paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents","version":1},"reference_index":18,"source":"pdf_text","source_observed_at":"2026-06-29T21:16:48.677159Z"},"links":{"citing_paper":"/paper/2605.26269"},"observation_digest":"sha256:2e2ea638e27aca7dc750f9413bd9315f368cc74449bd3cba427d126ee1baddcb","observation_id":"83303f67-6044-4557-bbc1-a789279c0763","resolution":{"observed_at":"2026-06-29T21:16:48.677159Z","resolver_source":null,"status":"unresolved"},"standing_notice":{"events":[],"reason":"canonical_work_link_unavailable","source_receipts":[],"state":"unavailable"}}],"paper":{"arxiv_id":"2605.26269","last_updated":"2026-05-25T18:53:22Z","latest_version":1,"primary_category":"cs.CR","snapshot_observed_at":"2026-08-15T00:50:18.626097Z","submitted_at":"2026-05-25T18:53:22Z","title":"AgentSecBench: Measuring Prompt Injection, Privacy Leakage, and Tool-Use Integrity in LLM Agents"},"reference_resolution":{"displayed":18,"state_counts":{"malformed_identifier":0,"metadata_mismatch":0,"parse_uncertain":0,"unresolved":18,"verified_exact":0,"verified_fuzzy":0},"total_outbound_references":18},"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":"2026-08-20T06:33:59.587034+00:00","source":"crossref"},{"observed_at":"2026-08-20T06:33:54.927442+00:00","source":"retraction_watch"}],"thesis":"As of 21 August 2026, this Paper Citation Record lists 18 of 18 outbound references and 2 inbound Pith citation observations for arXiv:2605.26269."}