{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:GYG2KGIROP3JZK6FWR7TANXTET","short_pith_number":"pith:GYG2KGIR","schema_version":"1.0","canonical_sha256":"360da5191173f69cabc5b47f3036f324dac403287e3f5233ce5b8e3c2674924e","source":{"kind":"arxiv","id":"2501.11621","version":1},"attestation_state":"computed","paper":{"title":"Trojan Detection Through Pattern Recognition for Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Matthew Yudin, Rauf Izmailov, Razvan Stefanescu, Vedant Bhasin","submitted_at":"2025-01-20T17:36:04Z","abstract_excerpt":"Trojan backdoors can be injected into large language models at various stages, including pretraining, fine-tuning, and in-context learning, posing a significant threat to the model's alignment. Due to the nature of causal language modeling, detecting these triggers is challenging given the vast search space. In this study, we propose a multistage framework for detecting Trojan triggers in large language models consisting of token filtration, trigger identification, and trigger verification. We discuss existing trigger identification methods and propose two variants of a black-box trigger inver"},"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":"2501.11621","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-20T17:36:04Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"3ac4c90cf16310df117a0126f06a12f443931c95a2f65d7010729cad3433112d","abstract_canon_sha256":"66dfec5de8c9d0471adc4f0055593136464e68a4255de98e714b63c05d12c630"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:03:11.648963Z","signature_b64":"2u11+Q7XKBS8KEgZLQOHW9oNcDBwVB/R1+p7gbpdtYQHYMBHYoVMBsm95SOjqFLeNHpeqdR7+LKMLHOftUJ1CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"360da5191173f69cabc5b47f3036f324dac403287e3f5233ce5b8e3c2674924e","last_reissued_at":"2026-07-05T10:03:11.648497Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:03:11.648497Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Trojan Detection Through Pattern Recognition for Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Matthew Yudin, Rauf Izmailov, Razvan Stefanescu, Vedant Bhasin","submitted_at":"2025-01-20T17:36:04Z","abstract_excerpt":"Trojan backdoors can be injected into large language models at various stages, including pretraining, fine-tuning, and in-context learning, posing a significant threat to the model's alignment. Due to the nature of causal language modeling, detecting these triggers is challenging given the vast search space. In this study, we propose a multistage framework for detecting Trojan triggers in large language models consisting of token filtration, trigger identification, and trigger verification. We discuss existing trigger identification methods and propose two variants of a black-box trigger inver"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.11621","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/2501.11621/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":"2501.11621","created_at":"2026-07-05T10:03:11.648557+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.11621v1","created_at":"2026-07-05T10:03:11.648557+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.11621","created_at":"2026-07-05T10:03:11.648557+00:00"},{"alias_kind":"pith_short_12","alias_value":"GYG2KGIROP3J","created_at":"2026-07-05T10:03:11.648557+00:00"},{"alias_kind":"pith_short_16","alias_value":"GYG2KGIROP3JZK6F","created_at":"2026-07-05T10:03:11.648557+00:00"},{"alias_kind":"pith_short_8","alias_value":"GYG2KGIR","created_at":"2026-07-05T10:03:11.648557+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GYG2KGIROP3JZK6FWR7TANXTET","json":"https://pith.science/pith/GYG2KGIROP3JZK6FWR7TANXTET.json","graph_json":"https://pith.science/api/pith-number/GYG2KGIROP3JZK6FWR7TANXTET/graph.json","events_json":"https://pith.science/api/pith-number/GYG2KGIROP3JZK6FWR7TANXTET/events.json","paper":"https://pith.science/paper/GYG2KGIR"},"agent_actions":{"view_html":"https://pith.science/pith/GYG2KGIROP3JZK6FWR7TANXTET","download_json":"https://pith.science/pith/GYG2KGIROP3JZK6FWR7TANXTET.json","view_paper":"https://pith.science/paper/GYG2KGIR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.11621&json=true","fetch_graph":"https://pith.science/api/pith-number/GYG2KGIROP3JZK6FWR7TANXTET/graph.json","fetch_events":"https://pith.science/api/pith-number/GYG2KGIROP3JZK6FWR7TANXTET/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GYG2KGIROP3JZK6FWR7TANXTET/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GYG2KGIROP3JZK6FWR7TANXTET/action/storage_attestation","attest_author":"https://pith.science/pith/GYG2KGIROP3JZK6FWR7TANXTET/action/author_attestation","sign_citation":"https://pith.science/pith/GYG2KGIROP3JZK6FWR7TANXTET/action/citation_signature","submit_replication":"https://pith.science/pith/GYG2KGIROP3JZK6FWR7TANXTET/action/replication_record"}},"created_at":"2026-07-05T10:03:11.648557+00:00","updated_at":"2026-07-05T10:03:11.648557+00:00"}