{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:GTXU26UY446VV75TM4YNS47BTG","short_pith_number":"pith:GTXU26UY","schema_version":"1.0","canonical_sha256":"34ef4d7a98e73d5affb36730d973e199a25f3322de1fb317ee4cbc78c3c2a539","source":{"kind":"arxiv","id":"2601.12359","version":1},"attestation_state":"computed","paper":{"title":"Zero-Shot Embedding Drift Detection: A Lightweight Defense Against Prompt Injections in LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CR","authors_text":"Akitsugu Tanaka, Anirudh Sekar, Arjun Damerla, Jasmine Zhang, Kevin Zhu, Mrinal Agarwal, Rachel Sharma","submitted_at":"2026-01-18T11:33:35Z","abstract_excerpt":"Prompt injection attacks have become an increasing vulnerability for LLM applications, where adversarial prompts exploit indirect input channels such as emails or user-generated content to circumvent alignment safeguards and induce harmful or unintended outputs. Despite advances in alignment, even state-of-the-art LLMs remain broadly vulnerable to adversarial prompts, underscoring the urgent need for robust, productive, and generalizable detection mechanisms beyond inefficient, model-specific patches. In this work, we propose Zero-Shot Embedding Drift Detection (ZEDD), a lightweight, low-engin"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2601.12359","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2026-01-18T11:33:35Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"5ca58a1789baaadec25c4bb7a16d3dffbdc8c34e23b2e1be63294e1cd6441b98","abstract_canon_sha256":"d1aac221567d0d7acee7f14f225d1b3729ed556e339989edab5dde3924c68e91"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-08T01:03:55.736771Z","signature_b64":"2uibV0W1uu3Zu3wMhZxmFhXkHqOO1Cn/KfsjgcSDtD7EHcxTTRN0w3SGrBDRwdiRhFSBOTDv2pEwmKcaBwo2Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"34ef4d7a98e73d5affb36730d973e199a25f3322de1fb317ee4cbc78c3c2a539","last_reissued_at":"2026-06-08T01:03:55.735862Z","signature_status":"signed_v1","first_computed_at":"2026-06-08T01:03:55.735862Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Zero-Shot Embedding Drift Detection: A Lightweight Defense Against Prompt Injections in LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CR","authors_text":"Akitsugu Tanaka, Anirudh Sekar, Arjun Damerla, Jasmine Zhang, Kevin Zhu, Mrinal Agarwal, Rachel Sharma","submitted_at":"2026-01-18T11:33:35Z","abstract_excerpt":"Prompt injection attacks have become an increasing vulnerability for LLM applications, where adversarial prompts exploit indirect input channels such as emails or user-generated content to circumvent alignment safeguards and induce harmful or unintended outputs. Despite advances in alignment, even state-of-the-art LLMs remain broadly vulnerable to adversarial prompts, underscoring the urgent need for robust, productive, and generalizable detection mechanisms beyond inefficient, model-specific patches. In this work, we propose Zero-Shot Embedding Drift Detection (ZEDD), a lightweight, low-engin"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2601.12359","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2601.12359/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2601.12359","created_at":"2026-06-08T01:03:55.735987+00:00"},{"alias_kind":"arxiv_version","alias_value":"2601.12359v1","created_at":"2026-06-08T01:03:55.735987+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2601.12359","created_at":"2026-06-08T01:03:55.735987+00:00"},{"alias_kind":"pith_short_12","alias_value":"GTXU26UY446V","created_at":"2026-06-08T01:03:55.735987+00:00"},{"alias_kind":"pith_short_16","alias_value":"GTXU26UY446VV75T","created_at":"2026-06-08T01:03:55.735987+00:00"},{"alias_kind":"pith_short_8","alias_value":"GTXU26UY","created_at":"2026-06-08T01:03:55.735987+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2606.07968","citing_title":"RecurGuard: Runtime Monitoring for Reasoning-Token Consumption Attacks","ref_index":21,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GTXU26UY446VV75TM4YNS47BTG","json":"https://pith.science/pith/GTXU26UY446VV75TM4YNS47BTG.json","graph_json":"https://pith.science/api/pith-number/GTXU26UY446VV75TM4YNS47BTG/graph.json","events_json":"https://pith.science/api/pith-number/GTXU26UY446VV75TM4YNS47BTG/events.json","paper":"https://pith.science/paper/GTXU26UY"},"agent_actions":{"view_html":"https://pith.science/pith/GTXU26UY446VV75TM4YNS47BTG","download_json":"https://pith.science/pith/GTXU26UY446VV75TM4YNS47BTG.json","view_paper":"https://pith.science/paper/GTXU26UY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2601.12359&json=true","fetch_graph":"https://pith.science/api/pith-number/GTXU26UY446VV75TM4YNS47BTG/graph.json","fetch_events":"https://pith.science/api/pith-number/GTXU26UY446VV75TM4YNS47BTG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GTXU26UY446VV75TM4YNS47BTG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GTXU26UY446VV75TM4YNS47BTG/action/storage_attestation","attest_author":"https://pith.science/pith/GTXU26UY446VV75TM4YNS47BTG/action/author_attestation","sign_citation":"https://pith.science/pith/GTXU26UY446VV75TM4YNS47BTG/action/citation_signature","submit_replication":"https://pith.science/pith/GTXU26UY446VV75TM4YNS47BTG/action/replication_record"}},"created_at":"2026-06-08T01:03:55.735987+00:00","updated_at":"2026-06-08T01:03:55.735987+00:00"}