{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:A7424GN73OGWJ4SDYJ7WSVQAHG","short_pith_number":"pith:A7424GN7","schema_version":"1.0","canonical_sha256":"07f9ae19bfdb8d64f243c27f6956003997d945589156acade690c75e0513c434","source":{"kind":"arxiv","id":"2401.10862","version":3},"attestation_state":"computed","paper":{"title":"Pruning for Protection: Increasing Jailbreak Resistance in Aligned LLMs Without Fine-Tuning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CR"],"primary_cat":"cs.LG","authors_text":"Adib Hasan, Alex Wang, Ileana Rugina","submitted_at":"2024-01-19T18:05:34Z","abstract_excerpt":"This paper investigates the impact of model compression on the way Large Language Models (LLMs) process prompts, particularly concerning jailbreak resistance. We show that moderate WANDA pruning can enhance resistance to jailbreaking attacks without fine-tuning, while maintaining performance on standard benchmarks. To systematically evaluate this safety enhancement, we introduce a dataset of 225 harmful tasks across five categories. Our analysis of LLaMA-2 Chat, Vicuna 1.3, and Mistral Instruct v0.2 reveals that pruning benefits correlate with initial model safety levels. We interpret these re"},"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":"2401.10862","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-01-19T18:05:34Z","cross_cats_sorted":["cs.AI","cs.CL","cs.CR"],"title_canon_sha256":"a46d96e8c78c8601a0336c54b6b9387edd1c4d001ebd597f376ea294e3c7c3b2","abstract_canon_sha256":"8a612e745d48013c61617b7d5854160940f4ddf6e71887f2c14cf6c98098769a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:28:47.210406Z","signature_b64":"D37EG0aZ8pivuVCNrsr8EhxX+jC7L8Fy0YMfAVws/ZVwp4ZHTrv065ZbybKQdJmavePRzKf2YVERz9sDX1KWDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"07f9ae19bfdb8d64f243c27f6956003997d945589156acade690c75e0513c434","last_reissued_at":"2026-07-05T09:28:47.209912Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:28:47.209912Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Pruning for Protection: Increasing Jailbreak Resistance in Aligned LLMs Without Fine-Tuning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CR"],"primary_cat":"cs.LG","authors_text":"Adib Hasan, Alex Wang, Ileana Rugina","submitted_at":"2024-01-19T18:05:34Z","abstract_excerpt":"This paper investigates the impact of model compression on the way Large Language Models (LLMs) process prompts, particularly concerning jailbreak resistance. We show that moderate WANDA pruning can enhance resistance to jailbreaking attacks without fine-tuning, while maintaining performance on standard benchmarks. To systematically evaluate this safety enhancement, we introduce a dataset of 225 harmful tasks across five categories. Our analysis of LLaMA-2 Chat, Vicuna 1.3, and Mistral Instruct v0.2 reveals that pruning benefits correlate with initial model safety levels. We interpret these re"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2401.10862","kind":"arxiv","version":3},"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/2401.10862/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":"2401.10862","created_at":"2026-07-05T09:28:47.209975+00:00"},{"alias_kind":"arxiv_version","alias_value":"2401.10862v3","created_at":"2026-07-05T09:28:47.209975+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.10862","created_at":"2026-07-05T09:28:47.209975+00:00"},{"alias_kind":"pith_short_12","alias_value":"A7424GN73OGW","created_at":"2026-07-05T09:28:47.209975+00:00"},{"alias_kind":"pith_short_16","alias_value":"A7424GN73OGWJ4SD","created_at":"2026-07-05T09:28:47.209975+00:00"},{"alias_kind":"pith_short_8","alias_value":"A7424GN7","created_at":"2026-07-05T09:28:47.209975+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2512.20856","citing_title":"NVIDIA Nemotron 3: Efficient and Open Intelligence","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/A7424GN73OGWJ4SDYJ7WSVQAHG","json":"https://pith.science/pith/A7424GN73OGWJ4SDYJ7WSVQAHG.json","graph_json":"https://pith.science/api/pith-number/A7424GN73OGWJ4SDYJ7WSVQAHG/graph.json","events_json":"https://pith.science/api/pith-number/A7424GN73OGWJ4SDYJ7WSVQAHG/events.json","paper":"https://pith.science/paper/A7424GN7"},"agent_actions":{"view_html":"https://pith.science/pith/A7424GN73OGWJ4SDYJ7WSVQAHG","download_json":"https://pith.science/pith/A7424GN73OGWJ4SDYJ7WSVQAHG.json","view_paper":"https://pith.science/paper/A7424GN7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2401.10862&json=true","fetch_graph":"https://pith.science/api/pith-number/A7424GN73OGWJ4SDYJ7WSVQAHG/graph.json","fetch_events":"https://pith.science/api/pith-number/A7424GN73OGWJ4SDYJ7WSVQAHG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/A7424GN73OGWJ4SDYJ7WSVQAHG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/A7424GN73OGWJ4SDYJ7WSVQAHG/action/storage_attestation","attest_author":"https://pith.science/pith/A7424GN73OGWJ4SDYJ7WSVQAHG/action/author_attestation","sign_citation":"https://pith.science/pith/A7424GN73OGWJ4SDYJ7WSVQAHG/action/citation_signature","submit_replication":"https://pith.science/pith/A7424GN73OGWJ4SDYJ7WSVQAHG/action/replication_record"}},"created_at":"2026-07-05T09:28:47.209975+00:00","updated_at":"2026-07-05T09:28:47.209975+00:00"}