{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:XZVCGHVRE5X5EATHCIEPM5HMIF","short_pith_number":"pith:XZVCGHVR","schema_version":"1.0","canonical_sha256":"be6a231eb1276fd202671208f674ec415ed7fda518dae06062155312a6c5432e","source":{"kind":"arxiv","id":"2312.17673","version":2},"attestation_state":"computed","paper":{"title":"Jatmo: Prompt Injection Defense by Task-Specific Finetuning","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CR","authors_text":"Basel Alomair, Chawin Sitawarin, David Wagner, Elizabeth Sun, Julien Piet, Maha Alrashed, Sizhe Chen, Zeming Wei","submitted_at":"2023-12-29T16:37:53Z","abstract_excerpt":"Large Language Models (LLMs) are attracting significant research attention due to their instruction-following abilities, allowing users and developers to leverage LLMs for a variety of tasks. However, LLMs are vulnerable to prompt-injection attacks: a class of attacks that hijack the model's instruction-following abilities, changing responses to prompts to undesired, possibly malicious ones. In this work, we introduce Jatmo, a method for generating task-specific models resilient to prompt-injection attacks. Jatmo leverages the fact that LLMs can only follow instructions once they have undergon"},"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":"2312.17673","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CR","submitted_at":"2023-12-29T16:37:53Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"d929329627863d9444c193159fdcd12affedadc6b3cdd3dbd1f0542368514492","abstract_canon_sha256":"88e73357092258c38b20c9122bde47eebf842fd5b0221b18e3d992e1c3b5a5f6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:31:36.387118Z","signature_b64":"JcCJJNylrKSbJF72jKd1p9enrnzDCijLx/gSAzvhURvk+OtTVzk4hwHhNarofHBFcptyeyADufJKLFVu2KYLDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"be6a231eb1276fd202671208f674ec415ed7fda518dae06062155312a6c5432e","last_reissued_at":"2026-07-05T07:31:36.386632Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:31:36.386632Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Jatmo: Prompt Injection Defense by Task-Specific Finetuning","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CR","authors_text":"Basel Alomair, Chawin Sitawarin, David Wagner, Elizabeth Sun, Julien Piet, Maha Alrashed, Sizhe Chen, Zeming Wei","submitted_at":"2023-12-29T16:37:53Z","abstract_excerpt":"Large Language Models (LLMs) are attracting significant research attention due to their instruction-following abilities, allowing users and developers to leverage LLMs for a variety of tasks. However, LLMs are vulnerable to prompt-injection attacks: a class of attacks that hijack the model's instruction-following abilities, changing responses to prompts to undesired, possibly malicious ones. In this work, we introduce Jatmo, a method for generating task-specific models resilient to prompt-injection attacks. Jatmo leverages the fact that LLMs can only follow instructions once they have undergon"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2312.17673","kind":"arxiv","version":2},"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/2312.17673/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":"2312.17673","created_at":"2026-07-05T07:31:36.386699+00:00"},{"alias_kind":"arxiv_version","alias_value":"2312.17673v2","created_at":"2026-07-05T07:31:36.386699+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2312.17673","created_at":"2026-07-05T07:31:36.386699+00:00"},{"alias_kind":"pith_short_12","alias_value":"XZVCGHVRE5X5","created_at":"2026-07-05T07:31:36.386699+00:00"},{"alias_kind":"pith_short_16","alias_value":"XZVCGHVRE5X5EATH","created_at":"2026-07-05T07:31:36.386699+00:00"},{"alias_kind":"pith_short_8","alias_value":"XZVCGHVR","created_at":"2026-07-05T07:31:36.386699+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2402.06922","citing_title":"Whispers in the Machine: Confidentiality in Agentic Systems","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2403.10559","citing_title":"Generative Models and Connected and Automated Vehicles: A Survey in Exploring the Intersection of Transportation and AI","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2502.05206","citing_title":"Safety at Scale: A Comprehensive Survey of Large Model and Agent Safety","ref_index":139,"is_internal_anchor":false},{"citing_arxiv_id":"2504.20472","citing_title":"Robustness via Referencing: Defending against Prompt Injection Attacks by Referencing the Executed Instruction","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2410.07283","citing_title":"Prompt Infection: LLM-to-LLM Prompt Injection within Multi-Agent Systems","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17562","citing_title":"SafeAgent: A Runtime Protection Architecture for Agentic Systems","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XZVCGHVRE5X5EATHCIEPM5HMIF","json":"https://pith.science/pith/XZVCGHVRE5X5EATHCIEPM5HMIF.json","graph_json":"https://pith.science/api/pith-number/XZVCGHVRE5X5EATHCIEPM5HMIF/graph.json","events_json":"https://pith.science/api/pith-number/XZVCGHVRE5X5EATHCIEPM5HMIF/events.json","paper":"https://pith.science/paper/XZVCGHVR"},"agent_actions":{"view_html":"https://pith.science/pith/XZVCGHVRE5X5EATHCIEPM5HMIF","download_json":"https://pith.science/pith/XZVCGHVRE5X5EATHCIEPM5HMIF.json","view_paper":"https://pith.science/paper/XZVCGHVR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2312.17673&json=true","fetch_graph":"https://pith.science/api/pith-number/XZVCGHVRE5X5EATHCIEPM5HMIF/graph.json","fetch_events":"https://pith.science/api/pith-number/XZVCGHVRE5X5EATHCIEPM5HMIF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XZVCGHVRE5X5EATHCIEPM5HMIF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XZVCGHVRE5X5EATHCIEPM5HMIF/action/storage_attestation","attest_author":"https://pith.science/pith/XZVCGHVRE5X5EATHCIEPM5HMIF/action/author_attestation","sign_citation":"https://pith.science/pith/XZVCGHVRE5X5EATHCIEPM5HMIF/action/citation_signature","submit_replication":"https://pith.science/pith/XZVCGHVRE5X5EATHCIEPM5HMIF/action/replication_record"}},"created_at":"2026-07-05T07:31:36.386699+00:00","updated_at":"2026-07-05T07:31:36.386699+00:00"}