{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:P6NJZYXOSKHLZEZT7SVO3DEB76","short_pith_number":"pith:P6NJZYXO","schema_version":"1.0","canonical_sha256":"7f9a9ce2ee928ebc9333fcaaed8c81ffaf52875fa68fec80158519d8a5a5b332","source":{"kind":"arxiv","id":"2503.15092","version":1},"attestation_state":"computed","paper":{"title":"Towards Understanding the Safety Boundaries of DeepSeek Models: Evaluation and Findings","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CR","authors_text":"Aishan Liu, Dacheng Tao, Deyue Zhang, Guangyi Zheng, Quanchen Zou, Wenxin Zhang, Xianglong Liu, Yongxin Huang, Zonghao Ying","submitted_at":"2025-03-19T10:44:37Z","abstract_excerpt":"This study presents the first comprehensive safety evaluation of the DeepSeek models, focusing on evaluating the safety risks associated with their generated content. Our evaluation encompasses DeepSeek's latest generation of large language models, multimodal large language models, and text-to-image models, systematically examining their performance regarding unsafe content generation. Notably, we developed a bilingual (Chinese-English) safety evaluation dataset tailored to Chinese sociocultural contexts, enabling a more thorough evaluation of the safety capabilities of Chinese-developed model"},"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":"2503.15092","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2025-03-19T10:44:37Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"fed345e8602472d82195486d84521e04e51adf38ff2952ec9d7e9dbba509b175","abstract_canon_sha256":"5ca8349c428df1652b4aa68589ed891e334eed1feeb3a3543f7de3e0f0a88d0f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:34:45.963174Z","signature_b64":"4kFPA9Y9DJq1EKfF+7+/o/wUEA26iER0G0CacWrrjq+oiddAxwClwLxAcZlv5wT/lJMq/dLfYT8IBwkYGPiDAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7f9a9ce2ee928ebc9333fcaaed8c81ffaf52875fa68fec80158519d8a5a5b332","last_reissued_at":"2026-07-05T10:34:45.962546Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:34:45.962546Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Towards Understanding the Safety Boundaries of DeepSeek Models: Evaluation and Findings","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CR","authors_text":"Aishan Liu, Dacheng Tao, Deyue Zhang, Guangyi Zheng, Quanchen Zou, Wenxin Zhang, Xianglong Liu, Yongxin Huang, Zonghao Ying","submitted_at":"2025-03-19T10:44:37Z","abstract_excerpt":"This study presents the first comprehensive safety evaluation of the DeepSeek models, focusing on evaluating the safety risks associated with their generated content. Our evaluation encompasses DeepSeek's latest generation of large language models, multimodal large language models, and text-to-image models, systematically examining their performance regarding unsafe content generation. Notably, we developed a bilingual (Chinese-English) safety evaluation dataset tailored to Chinese sociocultural contexts, enabling a more thorough evaluation of the safety capabilities of Chinese-developed model"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.15092","kind":"arxiv","version":1},"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/2503.15092/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":"2503.15092","created_at":"2026-07-05T10:34:45.962617+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.15092v1","created_at":"2026-07-05T10:34:45.962617+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.15092","created_at":"2026-07-05T10:34:45.962617+00:00"},{"alias_kind":"pith_short_12","alias_value":"P6NJZYXOSKHL","created_at":"2026-07-05T10:34:45.962617+00:00"},{"alias_kind":"pith_short_16","alias_value":"P6NJZYXOSKHLZEZT","created_at":"2026-07-05T10:34:45.962617+00:00"},{"alias_kind":"pith_short_8","alias_value":"P6NJZYXO","created_at":"2026-07-05T10:34:45.962617+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.24414","citing_title":"JT-SAFE-V2: Safety-by-Design Foundation Model with World-Context Data","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2507.21540","citing_title":"PRISM: Programmatic Reasoning with Image Sequence Manipulation for LVLM Jailbreaking","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2510.10073","citing_title":"SecureWebArena: A Holistic Security Evaluation Benchmark for LVLM-based Web Agents","ref_index":60,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08930","citing_title":"Internalizing Safety Understanding in Large Reasoning Models via Verification","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15725","citing_title":"Reasoning-targeted Jailbreak Attacks on Large Reasoning Models via Semantic Triggers and Psychological Framing","ref_index":48,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/P6NJZYXOSKHLZEZT7SVO3DEB76","json":"https://pith.science/pith/P6NJZYXOSKHLZEZT7SVO3DEB76.json","graph_json":"https://pith.science/api/pith-number/P6NJZYXOSKHLZEZT7SVO3DEB76/graph.json","events_json":"https://pith.science/api/pith-number/P6NJZYXOSKHLZEZT7SVO3DEB76/events.json","paper":"https://pith.science/paper/P6NJZYXO"},"agent_actions":{"view_html":"https://pith.science/pith/P6NJZYXOSKHLZEZT7SVO3DEB76","download_json":"https://pith.science/pith/P6NJZYXOSKHLZEZT7SVO3DEB76.json","view_paper":"https://pith.science/paper/P6NJZYXO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.15092&json=true","fetch_graph":"https://pith.science/api/pith-number/P6NJZYXOSKHLZEZT7SVO3DEB76/graph.json","fetch_events":"https://pith.science/api/pith-number/P6NJZYXOSKHLZEZT7SVO3DEB76/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/P6NJZYXOSKHLZEZT7SVO3DEB76/action/timestamp_anchor","attest_storage":"https://pith.science/pith/P6NJZYXOSKHLZEZT7SVO3DEB76/action/storage_attestation","attest_author":"https://pith.science/pith/P6NJZYXOSKHLZEZT7SVO3DEB76/action/author_attestation","sign_citation":"https://pith.science/pith/P6NJZYXOSKHLZEZT7SVO3DEB76/action/citation_signature","submit_replication":"https://pith.science/pith/P6NJZYXOSKHLZEZT7SVO3DEB76/action/replication_record"}},"created_at":"2026-07-05T10:34:45.962617+00:00","updated_at":"2026-07-05T10:34:45.962617+00:00"}