{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:H37IGTXUQIDDXXTF3M4QC43S4F","short_pith_number":"pith:H37IGTXU","schema_version":"1.0","canonical_sha256":"3efe834ef482063bde65db39017372e15aa7cc3e2e58cc62f1ca71f1e1629ced","source":{"kind":"arxiv","id":"2505.07608","version":2},"attestation_state":"computed","paper":{"title":"MiMo: Unlocking the Reasoning Potential of Language Model -- From Pretraining to Posttraining","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bowen Shen, Bowen Ye, Can Cai, Chenhong He, Cici, Dawei Zhu, Di Zhang, Dong Zhang, Duo Zhang, Gang Wang, Guoan Wang, Hailin Zhang, Haochen Zhao, Hao Tian, Heng Qu, Hongshen Xu, Huaqiu Liu, Jiebao Xiao, Jinhao Dong, Jun Shi, Kai Fang, Kainan Bao, Kangyang Zhou, Kang Zhou, Lei Li, Liang Zhao, LLM-Core Xiaomi: Bingquan Xia, Menghang Zhu, Nuo Chen, Peidian Li, Peng Wang, Qiantong Wang, Shaohui Liu, Shicheng Li, Shihua Yu, Shimao Chen, Shuhao Gu, Shuhuai Ren, Shuo Liu, Sirui Deng, Weiji Zhuang, Weikun Wang, Weiwei Lv, Wenhan Ma, Wenyu Yang, Xiangwei Deng, Xingchen Song, Xing Yong, Xing Zhang, Xin Zhang, Xinzhe Xu, Xu Wang, Yifan Song, Yihan Yan, Yi Huang, Yuanyuan Tian, Yudong Wang, Yue Yu, Yu Tu, Zhenru Lin, Zhichao Song, Zihan Jiang, Zihao Yue","submitted_at":"2025-05-12T14:30:11Z","abstract_excerpt":"We present MiMo-7B, a large language model born for reasoning tasks, with optimization across both pre-training and post-training stages. During pre-training, we enhance the data preprocessing pipeline and employ a three-stage data mixing strategy to strengthen the base model's reasoning potential. MiMo-7B-Base is pre-trained on 25 trillion tokens, with additional Multi-Token Prediction objective for enhanced performance and accelerated inference speed. During post-training, we curate a dataset of 130K verifiable mathematics and programming problems for reinforcement learning, integrating a te"},"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":"2505.07608","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-05-12T14:30:11Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"4e3320ae491ed6c4a6540aef866bd8cc583bb0b7a6a5e673b9e2a3e0bc20ac1f","abstract_canon_sha256":"4c41f96e17f9aaf831445ae95eefabd4ae6e75dc8d9490e11a5e8ed56157b6d6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:16:34.142342Z","signature_b64":"AuVF5lvE8BzaNIB5LATvoQiPDbg+DPSeNhxiafCfZLLcHpQtohfb5F3YgZKCJbdeKFVJoLIYW4jJE3Q7ezXVBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3efe834ef482063bde65db39017372e15aa7cc3e2e58cc62f1ca71f1e1629ced","last_reissued_at":"2026-07-05T11:16:34.141762Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:16:34.141762Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MiMo: Unlocking the Reasoning Potential of Language Model -- From Pretraining to Posttraining","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bowen Shen, Bowen Ye, Can Cai, Chenhong He, Cici, Dawei Zhu, Di Zhang, Dong Zhang, Duo Zhang, Gang Wang, Guoan Wang, Hailin Zhang, Haochen Zhao, Hao Tian, Heng Qu, Hongshen Xu, Huaqiu Liu, Jiebao Xiao, Jinhao Dong, Jun Shi, Kai Fang, Kainan Bao, Kangyang Zhou, Kang Zhou, Lei Li, Liang Zhao, LLM-Core Xiaomi: Bingquan Xia, Menghang Zhu, Nuo Chen, Peidian Li, Peng Wang, Qiantong Wang, Shaohui Liu, Shicheng Li, Shihua Yu, Shimao Chen, Shuhao Gu, Shuhuai Ren, Shuo Liu, Sirui Deng, Weiji Zhuang, Weikun Wang, Weiwei Lv, Wenhan Ma, Wenyu Yang, Xiangwei Deng, Xingchen Song, Xing Yong, Xing Zhang, Xin Zhang, Xinzhe Xu, Xu Wang, Yifan Song, Yihan Yan, Yi Huang, Yuanyuan Tian, Yudong Wang, Yue Yu, Yu Tu, Zhenru Lin, Zhichao Song, Zihan Jiang, Zihao Yue","submitted_at":"2025-05-12T14:30:11Z","abstract_excerpt":"We present MiMo-7B, a large language model born for reasoning tasks, with optimization across both pre-training and post-training stages. During pre-training, we enhance the data preprocessing pipeline and employ a three-stage data mixing strategy to strengthen the base model's reasoning potential. MiMo-7B-Base is pre-trained on 25 trillion tokens, with additional Multi-Token Prediction objective for enhanced performance and accelerated inference speed. During post-training, we curate a dataset of 130K verifiable mathematics and programming problems for reinforcement learning, integrating a te"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.07608","kind":"arxiv","version":2},"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/2505.07608/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":"2505.07608","created_at":"2026-07-05T11:16:34.141832+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.07608v2","created_at":"2026-07-05T11:16:34.141832+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.07608","created_at":"2026-07-05T11:16:34.141832+00:00"},{"alias_kind":"pith_short_12","alias_value":"H37IGTXUQIDD","created_at":"2026-07-05T11:16:34.141832+00:00"},{"alias_kind":"pith_short_16","alias_value":"H37IGTXUQIDDXXTF","created_at":"2026-07-05T11:16:34.141832+00:00"},{"alias_kind":"pith_short_8","alias_value":"H37IGTXU","created_at":"2026-07-05T11:16:34.141832+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":23,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.20280","citing_title":"ELVA: Exploring Ranking-Driven Universal Multimodal Retrieval","ref_index":62,"is_internal_anchor":false},{"citing_arxiv_id":"2606.19257","citing_title":"DreamReasoner-8B: Block-Size Curriculum Learning for Diffusion Reasoning Models","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2606.18388","citing_title":"LLMZero: Discovering Adaptive Training Strategies for RL Post-Training via LLM Agents","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10385","citing_title":"Beyond Absolute Imitation: Anchored Residual Guidance for Privileged On-Policy Distillation","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06453","citing_title":"Vortex: Efficient and Programmable Sparse Attention Serving for AI Agents","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05868","citing_title":"YouZhi: Towards High-Concurrency Financial LLMs via Adaptive GQA-to-MLA Transition","ref_index":60,"is_internal_anchor":false},{"citing_arxiv_id":"2606.04517","citing_title":"Treat Traffic Like Trees: A Semantic-Preserving Hierarchical Graph-Based Expert Framework for Encrypted Traffic Analysis","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2606.01249","citing_title":"Trust Region On-Policy Distillation","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28396","citing_title":"ADWIN: Adaptive Windows for Horizon-Aware On-Policy Distillation","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28077","citing_title":"MACReD: A Multi-Agent Collaborative Reasoning Framework for Reaction Diagram Parsing","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00750","citing_title":"I-WebGenBench : Evaluating Interactivity in LLM-Generated Scientific Web Applications","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2606.22974","citing_title":"When Preferences Fail to Become Incentives: A Utility-Behavior Gap in Large Language Models","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2504.12501","citing_title":"Reinforcement Learning from Human Feedback","ref_index":61,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06534","citing_title":"ROSE: Rollout On Serving GPUs via Cooperative Elasticity for Agentic RL","ref_index":77,"is_internal_anchor":false},{"citing_arxiv_id":"2509.15602","citing_title":"TennisTV: Do Multimodal Large Language Models Understand Tennis Rallies?","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2510.23477","citing_title":"MMTutorBench: The First Multimodal Benchmark for AI Math Tutoring","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10187","citing_title":"SciVQR: A Multidisciplinary Multimodal Benchmark for Advanced Scientific Reasoning Evaluation","ref_index":93,"is_internal_anchor":false},{"citing_arxiv_id":"2601.02780","citing_title":"MiMo-V2-Flash Technical Report","ref_index":49,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10187","citing_title":"SciVQR: A Multidisciplinary Multimodal Benchmark for Advanced Scientific Reasoning Evaluation","ref_index":93,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09934","citing_title":"TRACER: Verifiable Generative Provenance for Multimodal Tool-Using Agents","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03903","citing_title":"CC-OCR V2: Benchmarking Large Multimodal Models for Literacy in Real-world Document Processing","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06534","citing_title":"ROSE: Rollout On Serving GPUs via Cooperative Elasticity for Agentic RL","ref_index":78,"is_internal_anchor":false},{"citing_arxiv_id":"2604.22050","citing_title":"LayerBoost: Layer-Aware Attention Reduction for Efficient LLMs","ref_index":3,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/H37IGTXUQIDDXXTF3M4QC43S4F","json":"https://pith.science/pith/H37IGTXUQIDDXXTF3M4QC43S4F.json","graph_json":"https://pith.science/api/pith-number/H37IGTXUQIDDXXTF3M4QC43S4F/graph.json","events_json":"https://pith.science/api/pith-number/H37IGTXUQIDDXXTF3M4QC43S4F/events.json","paper":"https://pith.science/paper/H37IGTXU"},"agent_actions":{"view_html":"https://pith.science/pith/H37IGTXUQIDDXXTF3M4QC43S4F","download_json":"https://pith.science/pith/H37IGTXUQIDDXXTF3M4QC43S4F.json","view_paper":"https://pith.science/paper/H37IGTXU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.07608&json=true","fetch_graph":"https://pith.science/api/pith-number/H37IGTXUQIDDXXTF3M4QC43S4F/graph.json","fetch_events":"https://pith.science/api/pith-number/H37IGTXUQIDDXXTF3M4QC43S4F/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/H37IGTXUQIDDXXTF3M4QC43S4F/action/timestamp_anchor","attest_storage":"https://pith.science/pith/H37IGTXUQIDDXXTF3M4QC43S4F/action/storage_attestation","attest_author":"https://pith.science/pith/H37IGTXUQIDDXXTF3M4QC43S4F/action/author_attestation","sign_citation":"https://pith.science/pith/H37IGTXUQIDDXXTF3M4QC43S4F/action/citation_signature","submit_replication":"https://pith.science/pith/H37IGTXUQIDDXXTF3M4QC43S4F/action/replication_record"}},"created_at":"2026-07-05T11:16:34.141832+00:00","updated_at":"2026-07-05T11:16:34.141832+00:00"}