{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:NLU3PTFOQ7LLJGYUVLTPIYRYKD","short_pith_number":"pith:NLU3PTFO","schema_version":"1.0","canonical_sha256":"6ae9b7ccae87d6b49b14aae6f4623850cc498da61cafdf073d2df5c5e06c0181","source":{"kind":"arxiv","id":"2507.03236","version":2},"attestation_state":"computed","paper":{"title":"On Jailbreaking Quantized Language Models Through Fault Injection Attacks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Ahmad Tahmasivand, Ihsen Alouani, Khaled Khasawneh, Mohammed E. Fouda, Noureldin Zahran","submitted_at":"2025-07-04T00:48:48Z","abstract_excerpt":"The safety alignment of Language Models (LMs) is a critical concern, yet their integrity can be challenged by direct parameter manipulation attacks, such as those potentially induced by fault injection. As LMs are increasingly deployed using low-precision quantization for efficiency, this paper investigates the efficacy of such attacks for jailbreaking aligned LMs across different quantization schemes. We propose gradient-guided attacks, including a tailored progressive bit-level search algorithm introduced herein and a comparative word-level (single weight update) attack. Our evaluation on Ll"},"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":"2507.03236","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2025-07-04T00:48:48Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"e68e7cb4b36a7e0664c2f17e163d329ec3d36ddb069003ecb6fb4f03de8bc4b1","abstract_canon_sha256":"f93f3f4a2529cee8b26c751a03c431a31c34a1033ca11750da8baad9b6d53c7f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:34:11.207859Z","signature_b64":"+r2ZVbcKMuOTI8S61q1la86z4highWdmSNKEd9wbX0YUoCSKRIg9SaXXdgeFBKRgYEvuSHd9QaKk4dPlcFqeBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6ae9b7ccae87d6b49b14aae6f4623850cc498da61cafdf073d2df5c5e06c0181","last_reissued_at":"2026-07-05T11:34:11.207405Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:34:11.207405Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On Jailbreaking Quantized Language Models Through Fault Injection Attacks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Ahmad Tahmasivand, Ihsen Alouani, Khaled Khasawneh, Mohammed E. Fouda, Noureldin Zahran","submitted_at":"2025-07-04T00:48:48Z","abstract_excerpt":"The safety alignment of Language Models (LMs) is a critical concern, yet their integrity can be challenged by direct parameter manipulation attacks, such as those potentially induced by fault injection. As LMs are increasingly deployed using low-precision quantization for efficiency, this paper investigates the efficacy of such attacks for jailbreaking aligned LMs across different quantization schemes. We propose gradient-guided attacks, including a tailored progressive bit-level search algorithm introduced herein and a comparative word-level (single weight update) attack. Our evaluation on Ll"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.03236","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/2507.03236/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":"2507.03236","created_at":"2026-07-05T11:34:11.207466+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.03236v2","created_at":"2026-07-05T11:34:11.207466+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.03236","created_at":"2026-07-05T11:34:11.207466+00:00"},{"alias_kind":"pith_short_12","alias_value":"NLU3PTFOQ7LL","created_at":"2026-07-05T11:34:11.207466+00:00"},{"alias_kind":"pith_short_16","alias_value":"NLU3PTFOQ7LLJGYU","created_at":"2026-07-05T11:34:11.207466+00:00"},{"alias_kind":"pith_short_8","alias_value":"NLU3PTFO","created_at":"2026-07-05T11:34:11.207466+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/NLU3PTFOQ7LLJGYUVLTPIYRYKD","json":"https://pith.science/pith/NLU3PTFOQ7LLJGYUVLTPIYRYKD.json","graph_json":"https://pith.science/api/pith-number/NLU3PTFOQ7LLJGYUVLTPIYRYKD/graph.json","events_json":"https://pith.science/api/pith-number/NLU3PTFOQ7LLJGYUVLTPIYRYKD/events.json","paper":"https://pith.science/paper/NLU3PTFO"},"agent_actions":{"view_html":"https://pith.science/pith/NLU3PTFOQ7LLJGYUVLTPIYRYKD","download_json":"https://pith.science/pith/NLU3PTFOQ7LLJGYUVLTPIYRYKD.json","view_paper":"https://pith.science/paper/NLU3PTFO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.03236&json=true","fetch_graph":"https://pith.science/api/pith-number/NLU3PTFOQ7LLJGYUVLTPIYRYKD/graph.json","fetch_events":"https://pith.science/api/pith-number/NLU3PTFOQ7LLJGYUVLTPIYRYKD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NLU3PTFOQ7LLJGYUVLTPIYRYKD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NLU3PTFOQ7LLJGYUVLTPIYRYKD/action/storage_attestation","attest_author":"https://pith.science/pith/NLU3PTFOQ7LLJGYUVLTPIYRYKD/action/author_attestation","sign_citation":"https://pith.science/pith/NLU3PTFOQ7LLJGYUVLTPIYRYKD/action/citation_signature","submit_replication":"https://pith.science/pith/NLU3PTFOQ7LLJGYUVLTPIYRYKD/action/replication_record"}},"created_at":"2026-07-05T11:34:11.207466+00:00","updated_at":"2026-07-05T11:34:11.207466+00:00"}