{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:KR5QJBT5AHTBWJDYPUVVYX3AOP","short_pith_number":"pith:KR5QJBT5","schema_version":"1.0","canonical_sha256":"547b04867d01e61b24787d2b5c5f6073c3cb6514c05fca221e4c19d3f66019b7","source":{"kind":"arxiv","id":"2501.10639","version":3},"attestation_state":"computed","paper":{"title":"Latent-space adversarial training with post-aware calibration for defending large language models against jailbreak attacks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CR","authors_text":"Dongsheng Shi, Liang He, Linlin Wang, Xiaoling Wang, Xin Yi, Yue Li","submitted_at":"2025-01-18T02:57:12Z","abstract_excerpt":"Ensuring safety alignment is a critical requirement for large language models (LLMs), particularly given increasing deployment in real-world applications. Despite considerable advancements, LLMs remain susceptible to jailbreak attacks, which exploit system vulnerabilities to circumvent safety measures and elicit harmful or inappropriate outputs. Furthermore, while adversarial training-based defense methods have shown promise, a prevalent issue is the unintended over-defense behavior, wherein models excessively reject benign queries, significantly undermining their practical utility. To address"},"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":"2501.10639","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2025-01-18T02:57:12Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"8ff933a3d8234f3d899bc0ed2ddcd2ba110376945d7a945ea1663a0210bf2e94","abstract_canon_sha256":"d4608de395c2c33549d69f8ea4140e0a730ce7e6ee777941ac812274b057cdb0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:12:24.434437Z","signature_b64":"gnd3hcovYsSi6EuYjjv64Sw/FrYAW7mYKNhs80w8q5meTSasEa+vdkfFD+oXI1YxCcEZu1A1svFhBGTPEORzDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"547b04867d01e61b24787d2b5c5f6073c3cb6514c05fca221e4c19d3f66019b7","last_reissued_at":"2026-07-05T11:12:24.433872Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:12:24.433872Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Latent-space adversarial training with post-aware calibration for defending large language models against jailbreak attacks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CR","authors_text":"Dongsheng Shi, Liang He, Linlin Wang, Xiaoling Wang, Xin Yi, Yue Li","submitted_at":"2025-01-18T02:57:12Z","abstract_excerpt":"Ensuring safety alignment is a critical requirement for large language models (LLMs), particularly given increasing deployment in real-world applications. Despite considerable advancements, LLMs remain susceptible to jailbreak attacks, which exploit system vulnerabilities to circumvent safety measures and elicit harmful or inappropriate outputs. Furthermore, while adversarial training-based defense methods have shown promise, a prevalent issue is the unintended over-defense behavior, wherein models excessively reject benign queries, significantly undermining their practical utility. To address"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.10639","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/2501.10639/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":"2501.10639","created_at":"2026-07-05T11:12:24.433932+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.10639v3","created_at":"2026-07-05T11:12:24.433932+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.10639","created_at":"2026-07-05T11:12:24.433932+00:00"},{"alias_kind":"pith_short_12","alias_value":"KR5QJBT5AHTB","created_at":"2026-07-05T11:12:24.433932+00:00"},{"alias_kind":"pith_short_16","alias_value":"KR5QJBT5AHTBWJDY","created_at":"2026-07-05T11:12:24.433932+00:00"},{"alias_kind":"pith_short_8","alias_value":"KR5QJBT5","created_at":"2026-07-05T11:12:24.433932+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07918","citing_title":"Efficient Safety Alignment of Language Models via Latent Personality Traits","ref_index":52,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KR5QJBT5AHTBWJDYPUVVYX3AOP","json":"https://pith.science/pith/KR5QJBT5AHTBWJDYPUVVYX3AOP.json","graph_json":"https://pith.science/api/pith-number/KR5QJBT5AHTBWJDYPUVVYX3AOP/graph.json","events_json":"https://pith.science/api/pith-number/KR5QJBT5AHTBWJDYPUVVYX3AOP/events.json","paper":"https://pith.science/paper/KR5QJBT5"},"agent_actions":{"view_html":"https://pith.science/pith/KR5QJBT5AHTBWJDYPUVVYX3AOP","download_json":"https://pith.science/pith/KR5QJBT5AHTBWJDYPUVVYX3AOP.json","view_paper":"https://pith.science/paper/KR5QJBT5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.10639&json=true","fetch_graph":"https://pith.science/api/pith-number/KR5QJBT5AHTBWJDYPUVVYX3AOP/graph.json","fetch_events":"https://pith.science/api/pith-number/KR5QJBT5AHTBWJDYPUVVYX3AOP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KR5QJBT5AHTBWJDYPUVVYX3AOP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KR5QJBT5AHTBWJDYPUVVYX3AOP/action/storage_attestation","attest_author":"https://pith.science/pith/KR5QJBT5AHTBWJDYPUVVYX3AOP/action/author_attestation","sign_citation":"https://pith.science/pith/KR5QJBT5AHTBWJDYPUVVYX3AOP/action/citation_signature","submit_replication":"https://pith.science/pith/KR5QJBT5AHTBWJDYPUVVYX3AOP/action/replication_record"}},"created_at":"2026-07-05T11:12:24.433932+00:00","updated_at":"2026-07-05T11:12:24.433932+00:00"}