{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:OKW3KMRIJ3K24QBULBC7GQQSSL","short_pith_number":"pith:OKW3KMRI","schema_version":"1.0","canonical_sha256":"72adb532284ed5ae40345845f3421292e5c778944b06651d7e022c142b0aebdf","source":{"kind":"arxiv","id":"2507.18576","version":3},"attestation_state":"computed","paper":{"title":"SafeWork-R1: Coevolving Safety and Intelligence under the AI-45$^{\\circ}$ Law","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.CV"],"primary_cat":"cs.AI","authors_text":"Bangwei Liu, Bowen Zhou, Bo Zhang, Caoyuan Ma, Chaochao Lu, Chao Yang, Chen Qian, Chiyu Chen, Dan Ding, Dengke Deng, Dongrui Liu, Dongxing Shi, Fenghua Weng, Futing Wang, Guanxu Chen, Han Qi, Hao Li, Haotian Liang, Hefeng Zhou, Hong Huang, Huacan Liu, Jiachen Ma, Jiashu Qu, Jiaxuan Guo, Jie Cheng, Jie Li, Jingjing Qu, Jing Shao, Jinxuan Zhang, Juncheng Li, Kaichen Huang, Lijun Li, Lingjie Chen, Lingxiao Du, Lingyu Li, Lingyu Meng, Lujun Gui, Mingkang Chen, Muhao Wei, Pengyu Zhu, Qianxi He, Qiaosheng Zhang, Qihan Ren, Qihua Liu, Qingnan Ren, Qingyu Ren, Qingzi Zhu, Qitan Lv, Ruijun Ge, Ruofan Wang, Shanghai AI Lab: Yicheng Bao, Shanzhe Lei, Shiyang Huang, Shuai Shao, Shujie Wang, Sirui Chen, Tianle Gu, Wanying Qu, Wenjie Wang, Wenqi Shao, Wenwen Qu, Xiangtian Li, Xiaoshan Ding, Xiaoya Lu, Xiaoya Ma, Xiaoye Qu, Xiaoyu Wen, Xingcheng Xu, Xingge Qiao, Xinhao Song, Xinquan Chen, Xinshun Feng, Xin Song, Xin Wang, Xuan Tong, Xueyan Li, Xuhao Hu, Xuhong Wang, Yafu Li, Yajie Wang, Yang Yao, Yan Teng, Yanwei Fu, Yazhe Niu, Yexin Zhang, Yi Ding, Yige Wang, Yingchun Wang, Yingtong Xiong, Yinqiang Zheng, Yixu Wang, Yi Yu, Yizhuo Ding, Yuanfu Wang, Yuan Pu, Yubo Zhu, Yu Cheng, Yudong Lu, Yuenan Hou, Yuexiao Liu, Yue Yang, Yuezhang Peng, Yulei Ye, Yunhao Chen, Yuqi Wu, Yuxian Jiang, Yuxuan Gao, Yuyu Fan, Zhanhui Zhou, Zhenghao Lu, Zhenyun Yin, Zhichen Dong, Zhixuan Liu, Zhongtian Ma, Ziqi Miao, Zongkai Liu","submitted_at":"2025-07-24T16:49:19Z","abstract_excerpt":"We introduce SafeWork-R1, a cutting-edge multimodal reasoning model that demonstrates the coevolution of capabilities and safety. It is developed by our proposed SafeLadder framework, which incorporates large-scale, progressive, safety-oriented reinforcement learning post-training, supported by a suite of multi-principled verifiers. Unlike previous alignment methods such as RLHF that simply learn human preferences, SafeLadder enables SafeWork-R1 to develop intrinsic safety reasoning and self-reflection abilities, giving rise to safety `aha' moments. Notably, SafeWork-R1 achieves an average imp"},"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":"2507.18576","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2025-07-24T16:49:19Z","cross_cats_sorted":["cs.CL","cs.CV"],"title_canon_sha256":"f1675f5f68fe930c123bf626179790548d7c076ef57130564c811dcfb90d8dff","abstract_canon_sha256":"3a9140b886b2df8a22c2fe68de30224df8c1fa85b73f2c56253fe391c7497284"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:49:58.140976Z","signature_b64":"XqziW9K7tu11QSJGmabEDXvMy4tMNItfMLmAek0dlZdw2bp4iXukWpE+VDkrCB6E/5ZBZBxbAT309cdtlvPcAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"72adb532284ed5ae40345845f3421292e5c778944b06651d7e022c142b0aebdf","last_reissued_at":"2026-07-05T11:49:58.140502Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:49:58.140502Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SafeWork-R1: Coevolving Safety and Intelligence under the AI-45$^{\\circ}$ Law","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.CV"],"primary_cat":"cs.AI","authors_text":"Bangwei Liu, Bowen Zhou, Bo Zhang, Caoyuan Ma, Chaochao Lu, Chao Yang, Chen Qian, Chiyu Chen, Dan Ding, Dengke Deng, Dongrui Liu, Dongxing Shi, Fenghua Weng, Futing Wang, Guanxu Chen, Han Qi, Hao Li, Haotian Liang, Hefeng Zhou, Hong Huang, Huacan Liu, Jiachen Ma, Jiashu Qu, Jiaxuan Guo, Jie Cheng, Jie Li, Jingjing Qu, Jing Shao, Jinxuan Zhang, Juncheng Li, Kaichen Huang, Lijun Li, Lingjie Chen, Lingxiao Du, Lingyu Li, Lingyu Meng, Lujun Gui, Mingkang Chen, Muhao Wei, Pengyu Zhu, Qianxi He, Qiaosheng Zhang, Qihan Ren, Qihua Liu, Qingnan Ren, Qingyu Ren, Qingzi Zhu, Qitan Lv, Ruijun Ge, Ruofan Wang, Shanghai AI Lab: Yicheng Bao, Shanzhe Lei, Shiyang Huang, Shuai Shao, Shujie Wang, Sirui Chen, Tianle Gu, Wanying Qu, Wenjie Wang, Wenqi Shao, Wenwen Qu, Xiangtian Li, Xiaoshan Ding, Xiaoya Lu, Xiaoya Ma, Xiaoye Qu, Xiaoyu Wen, Xingcheng Xu, Xingge Qiao, Xinhao Song, Xinquan Chen, Xinshun Feng, Xin Song, Xin Wang, Xuan Tong, Xueyan Li, Xuhao Hu, Xuhong Wang, Yafu Li, Yajie Wang, Yang Yao, Yan Teng, Yanwei Fu, Yazhe Niu, Yexin Zhang, Yi Ding, Yige Wang, Yingchun Wang, Yingtong Xiong, Yinqiang Zheng, Yixu Wang, Yi Yu, Yizhuo Ding, Yuanfu Wang, Yuan Pu, Yubo Zhu, Yu Cheng, Yudong Lu, Yuenan Hou, Yuexiao Liu, Yue Yang, Yuezhang Peng, Yulei Ye, Yunhao Chen, Yuqi Wu, Yuxian Jiang, Yuxuan Gao, Yuyu Fan, Zhanhui Zhou, Zhenghao Lu, Zhenyun Yin, Zhichen Dong, Zhixuan Liu, Zhongtian Ma, Ziqi Miao, Zongkai Liu","submitted_at":"2025-07-24T16:49:19Z","abstract_excerpt":"We introduce SafeWork-R1, a cutting-edge multimodal reasoning model that demonstrates the coevolution of capabilities and safety. It is developed by our proposed SafeLadder framework, which incorporates large-scale, progressive, safety-oriented reinforcement learning post-training, supported by a suite of multi-principled verifiers. Unlike previous alignment methods such as RLHF that simply learn human preferences, SafeLadder enables SafeWork-R1 to develop intrinsic safety reasoning and self-reflection abilities, giving rise to safety `aha' moments. Notably, SafeWork-R1 achieves an average imp"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.18576","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/2507.18576/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":"2507.18576","created_at":"2026-07-05T11:49:58.140560+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.18576v3","created_at":"2026-07-05T11:49:58.140560+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.18576","created_at":"2026-07-05T11:49:58.140560+00:00"},{"alias_kind":"pith_short_12","alias_value":"OKW3KMRIJ3K2","created_at":"2026-07-05T11:49:58.140560+00:00"},{"alias_kind":"pith_short_16","alias_value":"OKW3KMRIJ3K24QBU","created_at":"2026-07-05T11:49:58.140560+00:00"},{"alias_kind":"pith_short_8","alias_value":"OKW3KMRI","created_at":"2026-07-05T11:49:58.140560+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.01602","citing_title":"ProWAFT: A ROMA-LPD Instance for Workload-Aware and Dynamic Fault Tolerance in FPGA-Based CNN Accelerators","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2607.01602","citing_title":"ProWAFT: A ROMA-LPD Instance for Workload-Aware and Dynamic Fault Tolerance in FPGA-Based CNN Accelerators","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30189","citing_title":"DAIN: Dynamic Agent-Based Interaction Network for Efficient and Collaborative Multimodal Reasoning","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2603.01589","citing_title":"SafeSci: Safety Evaluation of Large Language Models in Science Domains and Beyond","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01899","citing_title":"Disentangling Intent from Role: Adversarial Self-Play for Persona-Invariant Safety Alignment","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2604.05952","citing_title":"Towards Trustworthy Report Generation: A Deep Research Agent with Progressive Confidence Estimation and Calibration","ref_index":7,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OKW3KMRIJ3K24QBULBC7GQQSSL","json":"https://pith.science/pith/OKW3KMRIJ3K24QBULBC7GQQSSL.json","graph_json":"https://pith.science/api/pith-number/OKW3KMRIJ3K24QBULBC7GQQSSL/graph.json","events_json":"https://pith.science/api/pith-number/OKW3KMRIJ3K24QBULBC7GQQSSL/events.json","paper":"https://pith.science/paper/OKW3KMRI"},"agent_actions":{"view_html":"https://pith.science/pith/OKW3KMRIJ3K24QBULBC7GQQSSL","download_json":"https://pith.science/pith/OKW3KMRIJ3K24QBULBC7GQQSSL.json","view_paper":"https://pith.science/paper/OKW3KMRI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.18576&json=true","fetch_graph":"https://pith.science/api/pith-number/OKW3KMRIJ3K24QBULBC7GQQSSL/graph.json","fetch_events":"https://pith.science/api/pith-number/OKW3KMRIJ3K24QBULBC7GQQSSL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OKW3KMRIJ3K24QBULBC7GQQSSL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OKW3KMRIJ3K24QBULBC7GQQSSL/action/storage_attestation","attest_author":"https://pith.science/pith/OKW3KMRIJ3K24QBULBC7GQQSSL/action/author_attestation","sign_citation":"https://pith.science/pith/OKW3KMRIJ3K24QBULBC7GQQSSL/action/citation_signature","submit_replication":"https://pith.science/pith/OKW3KMRIJ3K24QBULBC7GQQSSL/action/replication_record"}},"created_at":"2026-07-05T11:49:58.140560+00:00","updated_at":"2026-07-05T11:49:58.140560+00:00"}