{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5UYVCOMKCOJTR2IEPGHVDVJA2R","short_pith_number":"pith:5UYVCOMK","schema_version":"1.0","canonical_sha256":"ed3151398a139338e904798f51d520d456352fb42bff9651815d594e3691770d","source":{"kind":"arxiv","id":"2504.10081","version":1},"attestation_state":"computed","paper":{"title":"RealSafe-R1: Safety-Aligned DeepSeek-R1 without Compromising Reasoning Capability","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Dongbai Li, Yao Huang, Yichi Zhang, Yinpeng Dong, Zhijie Deng, Zihao Zeng","submitted_at":"2025-04-14T10:26:37Z","abstract_excerpt":"Large Reasoning Models (LRMs), such as OpenAI o1 and DeepSeek-R1, have been rapidly progressing and achieving breakthrough performance on complex reasoning tasks such as mathematics and coding. However, the open-source R1 models have raised safety concerns in wide applications, such as the tendency to comply with malicious queries, which greatly impacts the utility of these powerful models in their applications. In this paper, we introduce RealSafe-R1 as safety-aligned versions of DeepSeek-R1 distilled models. To train these models, we construct a dataset of 15k safety-aware reasoning trajecto"},"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":"2504.10081","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.AI","submitted_at":"2025-04-14T10:26:37Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"55a60b41763db992cb31ec4842de49f6ced49827f042fa05729f2dd06badc305","abstract_canon_sha256":"056e032765f1a1e5a8dcca2421ee4f1cc2844430e197a39338c54ddab0058128"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:48:47.384696Z","signature_b64":"Q/1H7abuq+7lNWjG10R3ehjd9H5r9zR0QYyA8GpgN2QQUowtFbJS+2EsmfPXgpaMCjkTPvl6VfY6Q9FdM3qTAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ed3151398a139338e904798f51d520d456352fb42bff9651815d594e3691770d","last_reissued_at":"2026-07-05T10:48:47.384192Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:48:47.384192Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"RealSafe-R1: Safety-Aligned DeepSeek-R1 without Compromising Reasoning Capability","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Dongbai Li, Yao Huang, Yichi Zhang, Yinpeng Dong, Zhijie Deng, Zihao Zeng","submitted_at":"2025-04-14T10:26:37Z","abstract_excerpt":"Large Reasoning Models (LRMs), such as OpenAI o1 and DeepSeek-R1, have been rapidly progressing and achieving breakthrough performance on complex reasoning tasks such as mathematics and coding. However, the open-source R1 models have raised safety concerns in wide applications, such as the tendency to comply with malicious queries, which greatly impacts the utility of these powerful models in their applications. In this paper, we introduce RealSafe-R1 as safety-aligned versions of DeepSeek-R1 distilled models. To train these models, we construct a dataset of 15k safety-aware reasoning trajecto"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.10081","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/2504.10081/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":"2504.10081","created_at":"2026-07-05T10:48:47.384251+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.10081v1","created_at":"2026-07-05T10:48:47.384251+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.10081","created_at":"2026-07-05T10:48:47.384251+00:00"},{"alias_kind":"pith_short_12","alias_value":"5UYVCOMKCOJT","created_at":"2026-07-05T10:48:47.384251+00:00"},{"alias_kind":"pith_short_16","alias_value":"5UYVCOMKCOJTR2IE","created_at":"2026-07-05T10:48:47.384251+00:00"},{"alias_kind":"pith_short_8","alias_value":"5UYVCOMK","created_at":"2026-07-05T10:48:47.384251+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.31748","citing_title":"Addressing Over-Refusal in LLMs with Competing Rewards","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28146","citing_title":"Cybersecurity AI (CAI) Dataset","ref_index":72,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00686","citing_title":"Dialectics of Alignment: Harnessing Unsafe Knowledge for Dynamic Safety Routing","ref_index":82,"is_internal_anchor":false},{"citing_arxiv_id":"2508.04204","citing_title":"ReasoningGuard: Safeguarding Large Reasoning Models with Inference-time Safety Aha Moments","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08936","citing_title":"Self-ReSET: Learning to Self-Recover from Unsafe Reasoning Trajectories","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18946","citing_title":"Reasoning Structure Matters for Safety Alignment of Reasoning Models","ref_index":27,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5UYVCOMKCOJTR2IEPGHVDVJA2R","json":"https://pith.science/pith/5UYVCOMKCOJTR2IEPGHVDVJA2R.json","graph_json":"https://pith.science/api/pith-number/5UYVCOMKCOJTR2IEPGHVDVJA2R/graph.json","events_json":"https://pith.science/api/pith-number/5UYVCOMKCOJTR2IEPGHVDVJA2R/events.json","paper":"https://pith.science/paper/5UYVCOMK"},"agent_actions":{"view_html":"https://pith.science/pith/5UYVCOMKCOJTR2IEPGHVDVJA2R","download_json":"https://pith.science/pith/5UYVCOMKCOJTR2IEPGHVDVJA2R.json","view_paper":"https://pith.science/paper/5UYVCOMK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.10081&json=true","fetch_graph":"https://pith.science/api/pith-number/5UYVCOMKCOJTR2IEPGHVDVJA2R/graph.json","fetch_events":"https://pith.science/api/pith-number/5UYVCOMKCOJTR2IEPGHVDVJA2R/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5UYVCOMKCOJTR2IEPGHVDVJA2R/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5UYVCOMKCOJTR2IEPGHVDVJA2R/action/storage_attestation","attest_author":"https://pith.science/pith/5UYVCOMKCOJTR2IEPGHVDVJA2R/action/author_attestation","sign_citation":"https://pith.science/pith/5UYVCOMKCOJTR2IEPGHVDVJA2R/action/citation_signature","submit_replication":"https://pith.science/pith/5UYVCOMKCOJTR2IEPGHVDVJA2R/action/replication_record"}},"created_at":"2026-07-05T10:48:47.384251+00:00","updated_at":"2026-07-05T10:48:47.384251+00:00"}