{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:THMDJAIDA3PROWYF7NDSTZK6G6","short_pith_number":"pith:THMDJAID","schema_version":"1.0","canonical_sha256":"99d834810306df175b05fb4729e55e37b6082cb5e2db760866d2c988d0889ca7","source":{"kind":"arxiv","id":"2309.14348","version":3},"attestation_state":"computed","paper":{"title":"Defending Against Alignment-Breaking Attacks via Robustly Aligned LLM","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CR","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bochuan Cao, Jinghui Chen, Lu Lin, Yuanpu Cao","submitted_at":"2023-09-18T02:07:22Z","abstract_excerpt":"Recently, Large Language Models (LLMs) have made significant advancements and are now widely used across various domains. Unfortunately, there has been a rising concern that LLMs can be misused to generate harmful or malicious content. Though a line of research has focused on aligning LLMs with human values and preventing them from producing inappropriate content, such alignments are usually vulnerable and can be bypassed by alignment-breaking attacks via adversarially optimized or handcrafted jailbreaking prompts. In this work, we introduce a Robustly Aligned LLM (RA-LLM) to defend against po"},"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":"2309.14348","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-09-18T02:07:22Z","cross_cats_sorted":["cs.AI","cs.CR","cs.LG"],"title_canon_sha256":"9d3be5c6edbceeed06179311464977d4c8b126875aa7b157ca0a274128ed53b1","abstract_canon_sha256":"9fe552bedd5443865ca549fe797c085c2891793d22798310333fb041e02ac926"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:30:29.766931Z","signature_b64":"lh0Wy6CQi4RnvfEWfsnIgKVc30EZc8FmMSlif+I4DqswcDPaVlEzHcaqZD4ck2MWRiidELhY+1lkNYx6y7b7AA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"99d834810306df175b05fb4729e55e37b6082cb5e2db760866d2c988d0889ca7","last_reissued_at":"2026-07-05T08:30:29.766168Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:30:29.766168Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Defending Against Alignment-Breaking Attacks via Robustly Aligned LLM","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CR","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bochuan Cao, Jinghui Chen, Lu Lin, Yuanpu Cao","submitted_at":"2023-09-18T02:07:22Z","abstract_excerpt":"Recently, Large Language Models (LLMs) have made significant advancements and are now widely used across various domains. Unfortunately, there has been a rising concern that LLMs can be misused to generate harmful or malicious content. Though a line of research has focused on aligning LLMs with human values and preventing them from producing inappropriate content, such alignments are usually vulnerable and can be bypassed by alignment-breaking attacks via adversarially optimized or handcrafted jailbreaking prompts. In this work, we introduce a Robustly Aligned LLM (RA-LLM) to defend against po"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2309.14348","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/2309.14348/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":"2309.14348","created_at":"2026-07-05T08:30:29.766260+00:00"},{"alias_kind":"arxiv_version","alias_value":"2309.14348v3","created_at":"2026-07-05T08:30:29.766260+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2309.14348","created_at":"2026-07-05T08:30:29.766260+00:00"},{"alias_kind":"pith_short_12","alias_value":"THMDJAIDA3PR","created_at":"2026-07-05T08:30:29.766260+00:00"},{"alias_kind":"pith_short_16","alias_value":"THMDJAIDA3PROWYF","created_at":"2026-07-05T08:30:29.766260+00:00"},{"alias_kind":"pith_short_8","alias_value":"THMDJAID","created_at":"2026-07-05T08:30:29.766260+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.24552","citing_title":"Ellipsoid Control: A White-list Jailbreak Defense via Benign Latent Modeling","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2409.18169","citing_title":"Harmful Fine-tuning Attacks and Defenses for Large Language Models: A Survey","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2508.04204","citing_title":"ReasoningGuard: Safeguarding Large Reasoning Models with Inference-time Safety Aha Moments","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2407.04295","citing_title":"Jailbreak Attacks and Defenses Against Large Language Models: A Survey","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2604.09747","citing_title":"ADAM: A Systematic Data Extraction Attack on Agent Memory via Adaptive Querying","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04749","citing_title":"AI Trust OS -- A Continuous Governance Framework for Autonomous AI Observability and Zero-Trust Compliance in Enterprise Environments","ref_index":42,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/THMDJAIDA3PROWYF7NDSTZK6G6","json":"https://pith.science/pith/THMDJAIDA3PROWYF7NDSTZK6G6.json","graph_json":"https://pith.science/api/pith-number/THMDJAIDA3PROWYF7NDSTZK6G6/graph.json","events_json":"https://pith.science/api/pith-number/THMDJAIDA3PROWYF7NDSTZK6G6/events.json","paper":"https://pith.science/paper/THMDJAID"},"agent_actions":{"view_html":"https://pith.science/pith/THMDJAIDA3PROWYF7NDSTZK6G6","download_json":"https://pith.science/pith/THMDJAIDA3PROWYF7NDSTZK6G6.json","view_paper":"https://pith.science/paper/THMDJAID","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2309.14348&json=true","fetch_graph":"https://pith.science/api/pith-number/THMDJAIDA3PROWYF7NDSTZK6G6/graph.json","fetch_events":"https://pith.science/api/pith-number/THMDJAIDA3PROWYF7NDSTZK6G6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/THMDJAIDA3PROWYF7NDSTZK6G6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/THMDJAIDA3PROWYF7NDSTZK6G6/action/storage_attestation","attest_author":"https://pith.science/pith/THMDJAIDA3PROWYF7NDSTZK6G6/action/author_attestation","sign_citation":"https://pith.science/pith/THMDJAIDA3PROWYF7NDSTZK6G6/action/citation_signature","submit_replication":"https://pith.science/pith/THMDJAIDA3PROWYF7NDSTZK6G6/action/replication_record"}},"created_at":"2026-07-05T08:30:29.766260+00:00","updated_at":"2026-07-05T08:30:29.766260+00:00"}