---
title: "Baichuan-M3 — paper by Dou et al. (2026) — summary and ideas"
description: "Task-Specific RL (TaskRL) first builds each skill separately. 61 references, cited by 2 papers here."
authors: ["Chengfeng Dou", "Fan Yang", "Fei Li", "Jiyuan Jia", "Qiang Ju", "Shuai Wang", "Tianpeng Li", "Xiangrong Zeng", "Yijie Zhou", "Hongda Zhang", "Jinyang Tai", "Linzhuang Sun", "Peidong Guo", "Yichuan Mo", "Xiaochuan Wang", "Hengfu Cui", "Zhishou Zhang"]
year: 2026
venue: "arXiv:2602.06570"
source: "https://arxiv.org/abs/2602.06570"
doi: "10.48550/arxiv.2602.06570"
arxiv: "2602.06570"
licence: "CC BY 4.0 (arXiv)"
licence_url: "http://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/team-2026-m3-baichuan-m3-modeling-clinical/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/team-2026-m3-baichuan-m3-modeling-clinical/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/7/) / Baichuan-M3: Modeling Clinical Inquiry for Reliable Medical Decision-Making

# Baichuan-M3: Modeling Clinical Inquiry for Reliable Medical Decision-Making

**Authors:** [Chengfeng Dou](https://md.research.elamur.ai/free/en/authors/chengfeng-dou/), [Fan Yang](https://md.research.elamur.ai/free/en/authors/fan-yang/), [Fei Li](https://md.research.elamur.ai/free/en/authors/fei-li/), [Jiyuan Jia](https://md.research.elamur.ai/free/en/authors/jiyuan-jia/), [Qiang Ju](https://md.research.elamur.ai/free/en/authors/qiang-ju/), [Shuai Wang](https://md.research.elamur.ai/free/en/authors/shuai-wang/), [Tianpeng Li](https://md.research.elamur.ai/free/en/authors/tianpeng-li/), [Xiangrong Zeng](https://md.research.elamur.ai/free/en/authors/xiangrong-zeng/), [Yijie Zhou](https://md.research.elamur.ai/free/en/authors/yijie-zhou/), Hongda Zhang, Jinyang Tai, [Linzhuang Sun](https://md.research.elamur.ai/free/en/authors/linzhuang-sun/), Peidong Guo, Yichuan Mo, Xiaochuan Wang, Hengfu Cui and Zhishou Zhang · **Year:** 2026 · **Venue:** arXiv:2602.06570 · **Licence:** [CC BY 4.0 (arXiv)](http://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/2602.06570) · [PDF](https://arxiv.org/pdf/2602.06570) · [DOI 10.48550/arxiv.2602.06570](https://doi.org/10.48550/arxiv.2602.06570) · [arXiv 2602.06570](https://arxiv.org/abs/2602.06570) · [OpenAlex](https://openalex.org/W7128370680)

**Citations:** 0 citations (OpenAlex, 2026-10-14) · 2 in our library, of the 149,439 papers whose bibliographies we read

## Summary
Medical LLMs score well on static benchmarks but often stall in real consultations: they rarely ask for missing facts, and they may hallucinate when filling gaps. Baichuan-M3 is a medical-enhanced LLM trained to follow a physician's workflow: it gathers information actively, combines scattered evidence over long dialogues into a diagnosis, and suppresses hallucinations adaptively.

Training has three stages. Task-Specific RL (TaskRL) first builds each skill separately. Offline Policy Distillation and Multi-Teacher Online Policy Distillation (MOPD) then merge these skills into one policy. Segmented Pipeline RL gives separate rewards to inquiry, lab testing and diagnosis, which eases credit assignment in long dialogues. Dynamic Rubric Evolution counters reward saturation. Rewards come from a scripted patient simulator and a verification system that pairs rubric judging with search-backed fact checking.

The model reaches state-of-the-art results on HealthBench-Hard, the new HealthBench-Hallu and ScanBench, beating GPT-5.2. The authors argue that optimization aligned with clinical workflow is a practical route to clinical-grade assistants. The weights are released on Hugging Face.

## Key points
- **Benchmark results:** 44.4 on HealthBench-Hard. On ScanBench, an OSCE-style benchmark, the model scores 74.9 for Clinical Inquiry, 72.1 for Laboratory Testing and 74.4 for Diagnosis, ahead of GPT-5.2-High and expert baselines.
- **Patient simulator:** the simulated patient behaves passively. Most episodes (75%) start from an empty dialogue. The rest (25%) inject a dialogue snippet ending with a patient question; the physician agent sees the snippet, but the simulator does not.
- **Verify System:** it has two parts.
  - A Rubric Verifier: an LLM judge checks each rubric clause, and the clause weights, ranging from −10 to 10, are normalized into a reward between 0 and 1.
  - A Fact Verifier: responses are split into atomic claims, and each claim is checked through search against authoritative sources.
- **Faster inference:** Gated Eagle-3 adds Gated-Attention to the Eagle-3 draft model. This raises average acceptance length by 0.31 and throughput by 12%.
- **Quantization and limitations:** INT4 quantization uses self-generated calibration data (AutoRound, GPTQ format) so that all MoE experts are covered, and results are near-lossless. The model remains limited to episodic, text-only scenarios.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

📚 Shelves (5): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [👁️ Computer Vision and Pattern Recognition](https://md.research.elamur.ai/free/en/shelves/computer-science/computer-vision-and-pattern-recognition/) · 2,869 works · [🩺 Medicine](https://md.research.elamur.ai/free/en/shelves/medicine/) · 241 works · [🩺 Health Informatics](https://md.research.elamur.ai/free/en/shelves/medicine/health-informatics/) · 184 works

## Cite

Cite as: Chengfeng Dou, Fan Yang, Fei Li, Jiyuan Jia, Qiang Ju, Shuai Wang, Tianpeng Li, Xiangrong Zeng, Yijie Zhou, Hongda Zhang, Jinyang Tai, Linzhuang Sun, Peidong Guo, Yichuan Mo, Xiaochuan Wang, Hengfu Cui and Zhishou Zhang. 2026. Baichuan-M3: Modeling Clinical Inquiry for Reliable Medical Decision-Making. arXiv:2602.06570.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@misc{dou2026baichuan,
  title = {{Baichuan-M3: Modeling Clinical Inquiry for Reliable Medical Decision-Making}},
  author = {Chengfeng Dou, Fan Yang, Fei Li, Jiyuan Jia, Qiang Ju, Shuai Wang, Tianpeng Li, Xiangrong Zeng, Yijie Zhou, Hongda Zhang, Jinyang Tai, Linzhuang Sun, Peidong Guo, Yichuan Mo, Xiaochuan Wang, Hengfu Cui and Zhishou Zhang},
  year = {2026},
  doi = {10.48550/arxiv.2602.06570},
  eprint = {2602.06570},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2602.06570}
}
```

- [← previous AutoMonitor-Bench: Evaluating the Reliability of LLM-Based Misbehavior Monitor](https://md.research.elamur.ai/free/en/papers/yang-2026-automonitor-bench-evaluating-reliability-llm/)
- [next → BanglaLlama: LLaMA for Bangla Language](https://md.research.elamur.ai/free/en/papers/zehady-2026-banglallama-llama-bangla-language/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/7/)

## References

61 references, 24 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. OpenAI. [GPT-4 Technical Report](https://md.research.elamur.ai/free/en/papers/openai-2023-gpt-4-technical-report/). 2023. arXiv: 2303.08774. url: https://arxiv.org/abs/2303.08774. [↗](https://arxiv.org/abs/2303.08774)
2. Aaron Jaech, Adam Kalai, Adam Lerer, et al. “[OpenAI o1 System Card](https://md.research.elamur.ai/free/en/papers/openai-2024-openai-o1-system-card/)”. In: CoRR abs/2412.16720 (2024). doi: 10.48550/ARXIV.2412.16720. arXiv: 2412.16720. url: https://doi.org/10.48550/arXiv.2412.16720. [↗](https://doi.org/10.48550/arXiv.2412.16720)
3. Aiyuan Yang, Bin Xiao, Bingning Wang, et al. “[Baichuan 2: Open Large-scale Language Models](https://md.research.elamur.ai/free/en/papers/yang-2023-baichuan-2-open-large-scale/)”. In: CoRR abs/2309.10305 (2023). doi: 10.48550/ARXIV.2309.10305. arXiv: 2309.10305. url: https://doi.org/10.48550/arXiv.2309.10305. [↗](https://doi.org/10.48550/arXiv.2309.10305)
4. Yadong Li, Jun Liu, Tao Zhang, et al. “[Baichuan-Omni-1.5 Technical Report](https://md.research.elamur.ai/free/en/papers/li-2025-baichuan-omni-1-5/)”. In: CoRR abs/2501.15368 (2025). doi: 10.48550/ARXIV.2501.15368. arXiv: 2501.15368. url: https://doi.org/10.48550/arXiv.2501.15368. [↗](https://doi.org/10.48550/arXiv.2501.15368)
5. Bingning Wang, Haizhou Zhao, Huozhi Zhou, et al. “[Baichuan-M1: Pushing the Medical Capability of Large Language Models](https://md.research.elamur.ai/free/en/papers/wang-2025-baichuan-m1-pushing-medical/)”. In: CoRR abs/2502.12671 (2025). doi: 10.48550/ARXIV.2502.12671. arXiv: 2502.12671. url: https://doi.org/10.48550/arXiv.2502.12671. [↗](https://doi.org/10.48550/arXiv.2502.12671)
6. Chengfeng Dou, Chong Liu, Fan Yang, et al. “[Baichuan-M2: Scaling Medical Capability with Large Verifier System](https://md.research.elamur.ai/free/en/papers/team-2025-m2-baichuan-m2-scaling-medical/)”. In: CoRR abs/2509.02208 (2025). doi: 10.48550/ARXIV.2509.02208. arXiv: 2509.02208. url: https://doi.org/10.48550/arXiv.2509.02208. [↗](https://doi.org/10.48550/arXiv.2509.02208)
7. Andrew Sellergren, Sahar Kazemzadeh, Tiam Jaroensri, et al. “MedGemma Technical Report”. In: CoRR abs/2507.05201 (2025). doi: 10.48550/ARXIV.2507.05201. arXiv: 2507.05201. url: https://doi.org/10.48550/arXiv.2507.05201. [↗](https://doi.org/10.48550/arXiv.2507.05201)
8. LASA Team, Weiwen Xu, Hou Pong Chan, et al. “Lingshu: A Generalist Foundation Model for Unified Multimodal Medical Understanding and Reasoning”. In: CoRR abs/2506.07044 (2025). doi: 10.48550/ARXIV.2506.07044. arXiv: 2506.07044. url: https://doi.org/10.48550/arXiv.2506.07044. — [doi:10.48550/arxiv.2506.07044](https://doi.org/10.48550/arxiv.2506.07044) [↗](https://doi.org/10.48550/arXiv.2506.07044)
9. Songtao Jiang, Yuan Wang, Sibo Song, et al. “Hulu-Med: A Transparent Generalist Model towards Holistic Medical Vision-Language Understanding”. In: CoRR abs/2510.08668 (2025). doi: 10.48550/ARXIV.2510.08668. arXiv: 2510.08668. url: https://doi.org/10.48550/arXiv.2510.08668. [↗](https://doi.org/10.48550/arXiv.2510.08668)
10. Binbin Li, Tianxin Meng, Xiaoming Shi, Jie Zhai, and Tong Ruan. “MedDM: LLM-executable clinical guidance tree for clinical decision-making”. In: CoRR abs/2312.02441 (2023). doi: 10.48550/ARXIV.2312.02441. arXiv: 2312.02441. url: https://doi.org/10.48550/arXiv.2312.02441. — [doi:10.48550/arxiv.2312.02441](https://doi.org/10.48550/arxiv.2312.02441) [↗](https://doi.org/10.48550/arXiv.2312.02441)
11. Shengxin Hong, Liang Xiao, Xin Zhang, and Jianxia Chen. “ArgMed-Agents: Explainable Clinical Decision Reasoning with LLM Disscusion via Argumentation Schemes”. In: IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2024, Lisbon, Portugal, December 3-6, 2024. Ed. by Mario Cannataro… — [doi:10.1109/bibm62325.2024.10822109](https://doi.org/10.1109/bibm62325.2024.10822109) [find](https://scholar.google.com/scholar?q=Shengxin+Hong%2C+Liang+Xiao%2C+Xin+Zhang%2C+and+Jianxia+Chen%2E+%E2%80%9CArgMed%2DAgents%3A+Explainable+Clinical+Decision+Reasoning+with+LLM+Disscusion+via+Argumentation+Schemes%E2%80%9D%2E+In%3A+IEEE+International+Conference+on+Bio)
12. OpenAI. Introducing GPT-5.2. [https://openai.com/index/introducing-gpt-5-2/](https://openai.com/index/introducing-gpt-5-2/). Accessed: 2026-01-29. Dec. 2025. [↗](https://openai.com/index/introducing-gpt-5-2/)
13. OpenAI. Introducing ChatGPT Health. Jan. 2026. url: [https://openai.com/index/introducing-chatgpt-health](https://openai.com/index/introducing-chatgpt-health). [↗](https://openai.com/index/introducing-chatgpt-health)
14. Anthropic. Advancing Claude in healthcare and the life sciences. Jan. 2026. url: [https://www.anthropic.com/news/healthcare-life-sciences](https://www.anthropic.com/news/healthcare-life-sciences). [↗](https://www.anthropic.com/news/healthcare-life-sciences)
15. Jia-Yu Yao, Kun-Peng Ning, Zhen-Hui Liu, Munan Ning, and Li Yuan. “LLM Lies: Hallucinations are not Bugs, but Features as Adversarial Examples”. In: CoRR abs/2310.01469 (2023). doi: 10.48550/ARXIV.2310.01469. arXiv: 2310.01469. url: https://doi.org/10.48550/arXiv.2310.01469. — [doi:10.48550/arxiv.2310.01469](https://doi.org/10.48550/arxiv.2310.01469) [↗](https://doi.org/10.48550/arXiv.2310.01469)
16. Rahul K. Arora, Jason Wei, Rebecca Soskin Hicks, et al. “[HealthBench: Evaluating Large Language Models Towards Improved Human Health](https://md.research.elamur.ai/free/en/papers/arora-2025-healthbench-evaluating-large-language-models/)”. In: CoRR abs/2505.08775 (2025). doi: 10.48550/ARXIV.2505.08775. arXiv: 2505.08775. url: https://doi.org/10.48550/arXiv.2505.08775. [↗](https://doi.org/10.48550/arXiv.2505.08775)
17. Ankit Pal, Logesh Kumar Umapathi, and Malaikannan Sankarasubbu. “[Med-HALT: Medical Domain Hallucination Test for Large Language Models](https://md.research.elamur.ai/free/en/papers/pal-2023-med-halt-medical-domain-hallucination/)”. In: Proceedings of the 27th Conference on Computational Natural Language Learning, CoNLL 2023, Singapore, December 6-7, 2023. Ed. by Jing Jiang, David Reitter, and…
18. Tao Tu, Anil Palepu, Mike Schaekermann, et al. “Towards Conversational Diagnostic AI”. In: CoRR abs/2401.05654 (2024). doi: 10.48550/ARXIV.2401.05654. arXiv: 2401.05654. url: https://doi.org/10.48550/arXiv.2401.05654. — [doi:10.48550/arxiv.2401.05654](https://doi.org/10.48550/arxiv.2401.05654) [↗](https://doi.org/10.48550/arXiv.2401.05654)
19. Elahe Vedadi, David G. T. Barrett, Natalie Harris, et al. “Towards physician-centered oversight of conversational diagnostic AI”. In: CoRR abs/2507.15743 (2025). doi: 10.48550/ARXIV.2507.15743. arXiv: 2507.15743. url: https://doi.org/10.48550/arXiv.2507.15743. — [doi:10.48550/arxiv.2507.15743](https://doi.org/10.48550/arxiv.2507.15743) [↗](https://doi.org/10.48550/arXiv.2507.15743)
20. Daya Guo, Dejian Yang, Haowei Zhang, et al. “DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning”. In: Nat. 645.8081 [(2025), pp](https://md.research.elamur.ai/free/en/papers/deepseek-ai-2025-deepseek-r1-incentivizing-reasoning-capability/). 633–638. doi: 10.1038/S41586-025-09422-Z. url: https://doi.org/10.1038/s41586-025-09422-z. [↗](https://doi.org/10.1038/s41586-025-09422-z)
21. Mingan Lin, Fan Yang, Yanjun Shen, et al. “[Baichuan Alignment Technical Report](https://md.research.elamur.ai/free/en/papers/lin-2024-baichuan-alignment-technical-report/)”. In: CoRR abs/2410.14940 (2024). doi: 10.48550/ARXIV.2410.14940. arXiv: 2410.14940. url: https://doi.org/10.48550/arXiv.2410.14940. [↗](https://doi.org/10.48550/arXiv.2410.14940)
22. LiChun Cao and ZhiMin. “An Overview of Deep Reinforcement Learning”. In: Proceedings of the 2019 4th International Conference on Automation, Control and Robotics Engineering, CACRE 2019, Shenzhen, China, July 19-21, 2019. ACM, 2019, 17:1–17:9. doi: 10.1145/3351917.3351989. url: https://doi.org/10.11… — [doi:10.1145/3351917.3351989](https://doi.org/10.1145/3351917.3351989) [find](https://scholar.google.com/scholar?q=LiChun+Cao+and+ZhiMin%2E+%E2%80%9CAn+Overview+of+Deep+Reinforcement+Learning%E2%80%9D%2E+In%3A+Proceedings+of+the+2019+4th+International+Conference+on+Automation%2C+Control+and+Robotics+Engineering%2C+CACRE+2019%2C+Shenzhen%2C+Chi)
23. Zijun Yao, Yantao Liu, Yanxu Chen, et al. “Are Reasoning Models More Prone to Hallucination?” In: CoRR abs/2505.23646 (2025). doi: 10.48550/ARXIV.2505.23646. arXiv: 2505.23646. url: https://doi.org/10.48550/arXiv.2505.23646. — [doi:10.48550/arxiv.2505.23646](https://doi.org/10.48550/arxiv.2505.23646) [↗](https://doi.org/10.48550/arXiv.2505.23646)
24. Chenlong Yin, Zeyang Sha, Shiwen Cui, and Changhua Meng. “[The Reasoning Trap: How Enhancing LLM Reasoning Amplifies Tool Hallucination](https://md.research.elamur.ai/free/en/papers/yin-2026-reasoning-trap-how-enhancing-llm/)”. In: CoRR abs/2510.22977 (2025). doi: 10.48550/ARXIV.2510.22977. arXiv: 2510.22977. url: https://doi.org/10.48550/arXiv.2510.22977. [↗](https://doi.org/10.48550/arXiv.2510.22977)
25. Zenan Huang, Yihong Zhuang, Guoshan Lu, et al. “Reinforcement Learning with Rubric Anchors”. In: CoRR abs/2508.12790 (2025). doi: 10.48550/ARXIV.2508.12790. arXiv: 2508.12790. url: https://doi.org/10.48550/arXiv.2508.12790. — [doi:10.48550/arxiv.2508.12790](https://doi.org/10.48550/arxiv.2508.12790) [↗](https://doi.org/10.48550/arXiv.2508.12790)
26. Anisha Gunjal, Anthony Wang, Elaine Lau, Vaskar Nath, Bing Liu, and Sean Hendryx. “Rubrics as Rewards: Reinforcement Learning Beyond Verifiable Domains”. In: CoRR abs/2507.17746 (2025). doi: 10.48550/ARXIV.2507.17746. arXiv: 2507.17746. url: https://doi.org/10.48550/arXiv.2507.17746. [↗](https://doi.org/10.48550/arXiv.2507.17746)
27. Jiawei Gu, Xuhui Jiang, Zhichao Shi, et al. “A Survey on LLM-as-a-Judge”. In: CoRR abs/2411.15594 (2024). doi: 10.48550/ARXIV.2411.15594. arXiv: 2411.15594. url: https://doi.org/10.48550/arXiv.2411.15594. [↗](https://doi.org/10.48550/arXiv.2411.15594)
28. Haoyang Li, Yiming Li, Anxin Tian, et al. “A Survey on Large Language Model Acceleration based on KV Cache Management”. In: Trans. Mach. Learn. Res. 2025 (2025). url: [https://openreview.net/forum?id=z3JZzu9EA3](https://openreview.net/forum?id=z3JZzu9EA3). [↗](https://openreview.net/forum?id=z3JZzu9EA3)
29. Sewon Min, Kalpesh Krishna, Xinxi Lyu, et al. “[FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation](https://md.research.elamur.ai/free/en/papers/min-2023-factscore-fine-grained-atomic-evaluation/)”. In: Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, EMNLP 2023, Singapore, December 6-10, 2023. Ed. by Houda Bouamor, J…
30. Jerry Wei, Chengrun Yang, Xinying Song, et al. “[Long-form factuality in large language models](https://md.research.elamur.ai/free/en/papers/wei-2024-long-form-factuality-large-language/)”. In: Advances in Neural Information Processing Systems 38: Annual Conference on Neural Information Processing Systems 2024, NeurIPS 2024, Vancouver, BC, Canada, December 10 - 15, 2024. Ed. by Amir Globerso…
31. Aaditya Singh, Adam Fry, Adam Perelman, et al. “[Openai gpt-5 system card](https://md.research.elamur.ai/free/en/papers/singh-2025-openai-gpt-5-system-card/)”. In: CoRR (2025). arXiv: 2601.03267.
32. Sachin Ravi, Sebastian Musslick, Maia Hamin, Theodore L. Willke, and Jonathan D. Cohen. “Navigating the Trade-Off between Multi-Task Learning and Learning to Multitask in Deep Neural Networks”. In: CoRR abs/2007.10527 (2020). arXiv: 2007.10527. url: https://arxiv.org/abs/2007.10527. — [arXiv:2007.10527](https://arxiv.org/abs/2007.10527) [↗](https://arxiv.org/abs/2007.10527)
33. Yichuan Mo and Shilin Wang. “Multi-Task Learning Improves Synthetic Speech Detection”. In: IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2022, Virtual and Singapore, 23-27 May 2022. IEEE, 2022, pp. 6392–6396. doi: 10.1109/ICASSP43922.2022.9746059. url: [https://doi.…](https://doi.…) — [doi:10.1109/icassp43922.2022.9746059](https://doi.org/10.1109/icassp43922.2022.9746059) [↗](https://doi.…)
34. Yisen Wang, Yichuan Mo, Hongjun Wang, Junyi Li, and Zhouchen Lin. “Generalist++: A Meta-learning Framework for Mitigating Trade-off in Adversarial Training”. In: CoRR abs/2510.13361 (2025). doi: 10.48550/ARXIV.2510.13361. arXiv: 2510.13361. url: https://doi.org/10.48550/arXiv.2510.13361. — [doi:10.48550/arxiv.2510.13361](https://doi.org/10.48550/arxiv.2510.13361) [↗](https://doi.org/10.48550/arXiv.2510.13361)
35. LLM-Core Xiaomi. [MiMo-V2-Flash Technical Report](https://md.research.elamur.ai/free/en/papers/team-2026-llm-mimo-v2-flash-technical/). 2025. url: [https://github.com/XiaomiMiMo/MiMo-V2-Flash/paper.pdf](https://github.com/XiaomiMiMo/MiMo-V2-Flash/paper.pdf). [↗](https://github.com/XiaomiMiMo/MiMo-V2-Flash/paper.pdf)
36. Kevin Lu and Thinking Machines Lab. “On-Policy Distillation”. In: Thinking Machines Lab: Connectionism (2025). [https://thinkingmachines.ai/blog/on-policy-distillation](https://thinkingmachines.ai/blog/on-policy-distillation). doi: 10.64434/tml.20251026. — [doi:10.64434/tml.20251026](https://doi.org/10.64434/tml.20251026) [↗](https://thinkingmachines.ai/blog/on-policy-distillation)
37. Jie Liu, Wenxuan Wang, Zizhan Ma, et al. Medchain: Bridging the Gap Between LLM Agents and Clinical Practice with Interactive Sequence. 2025. arXiv: 2412.01605 [cs.CL]. url: https://arxiv.org/abs/2412.01605. — [arXiv:2412.01605](https://arxiv.org/abs/2412.01605) [↗](https://arxiv.org/abs/2412.01605)
38. Kangenbei Liao, Qianlong Liu, Zhongyu Wei, et al. “Task-oriented Dialogue System for Automatic Disease Diagnosis via Hierarchical Reinforcement Learning”. In: CoRR abs/2004.14254 (2020). arXiv: 2004.14254. url: https://arxiv.org/abs/2004.14254. — [arXiv:2004.14254](https://arxiv.org/abs/2004.14254) [↗](https://arxiv.org/abs/2004.14254)
39. Chengfeng Dou, Ying Zhang, Zhi Jin, et al. “Integrating Physician Diagnostic Logic into Large Language Models: Preference Learning from Process Feedback”. In: Findings of the Association for Computational Linguistics, ACL 2024, Bangkok, Thailand and virtual meeting, August 11-16, 2024. Ed. by Lun-We… [find](https://scholar.google.com/scholar?q=Chengfeng+Dou%2C+Ying+Zhang%2C+Zhi+Jin%2C+et+al%2E+%E2%80%9CIntegrating+Physician+Diagnostic+Logic+into+Large+Language+Models%3A+Preference+Learning+from+Process+Feedback%E2%80%9D%2E+In%3A+Findings+of+the+Association+for+Computati)
40. Shihui Yang, Chengfeng Dou, Peidong Guo, et al. “DCPO: Dynamic Clipping Policy Optimization”. In: CoRR abs/2509.02333 (2025). doi: 10.48550/ARXIV.2509.02333. arXiv: 2509.02333. url: https://doi.org/10.48550/arXiv.2509.02333. — [doi:10.48550/arxiv.2509.02333](https://doi.org/10.48550/arxiv.2509.02333) [↗](https://doi.org/10.48550/arXiv.2509.02333)
41. Renrui Zhang, Chengzhuo Tong, Zhizheng Zhao, et al. “Let’s Verify and Reinforce Image Generation Step by Step”. In: IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2025, Nashville, TN, USA, June 11-15, 2025. Computer Vision Foundation / IEEE, 2025, pp. 28662–28672. doi: 10.1109/… — [doi:10.1109/cvpr52734.2025.02669](https://doi.org/10.1109/cvpr52734.2025.02669) [find](https://scholar.google.com/scholar?q=Renrui+Zhang%2C+Chengzhuo+Tong%2C+Zhizheng+Zhao%2C+et+al%2E+%E2%80%9CLet%E2%80%99s+Verify+and+Reinforce+Image+Generation+Step+by+Step%E2%80%9D%2E+In%3A+IEEE%2FCVF+Conference+on+Computer+Vision+and+Pattern+Recognition%2C+CVPR+2025%2C+Nashville)
42. Chujie Zheng, Shixuan Liu, Mingze Li, et al. “[Group Sequence Policy Optimization](https://md.research.elamur.ai/free/en/papers/zheng-2025-group-sequence-policy-optimization/)”. In: CoRR abs/2507.18071 (2025). doi: 10.48550/ARXIV.2507.18071. arXiv: 2507.18071. url: https://doi.org/10.48550/arXiv.2507.18071. [↗](https://doi.org/10.48550/arXiv.2507.18071)
43. Yoshua Bengio, Jérôme Louradour, Ronan Collobert, and Jason Weston. “[Curriculum learning](https://md.research.elamur.ai/free/en/papers/bengio-2009-curriculum-learning/)”. In: Proceedings of the 26th annual international conference on machine learning. 2009, pp. 41–48.
44. Rulin Shao, Akari Asai, Shannon Zejiang Shen, et al. “[DR Tulu: Reinforcement Learning with Evolving Rubrics for Deep Research](https://md.research.elamur.ai/free/en/papers/shao-2025-dr-tulu-reinforcement-learning-evolving/)”. In: CoRR abs/2511.19399 (2025). doi: 10.48550/ARXIV.2511.19399. arXiv: 2511.19399. url: https://doi.org/10.48550/arXiv.2511.19399. [↗](https://doi.org/10.48550/arXiv.2511.19399)
45. Junkai Zhang, Zihao Wang, Lin Gui, et al. “Chasing the Tail: Effective Rubric-based Reward Modeling for Large Language Model Post-Training”. In: CoRR abs/2509.21500 (2025). doi: 10.48550/ARXIV.2509.21500. arXiv: 2509.21500. url: https://doi.org/10.48550/arXiv.2509.21500. — [doi:10.48550/arxiv.2509.21500](https://doi.org/10.48550/arxiv.2509.21500) [↗](https://doi.org/10.48550/arXiv.2509.21500)
46. Xilun Chen, Ilia Kulikov, Vincent-Pierre Berges, et al. “Learning to Reason for Factuality”. In: CoRR abs/2508.05618 (2025). doi: 10.48550/ARXIV.2508.05618. arXiv: 2508.05618. url: https://doi.org/10.48550/arXiv.2508.05618. — [doi:10.48550/arxiv.2508.05618](https://doi.org/10.48550/arxiv.2508.05618) [↗](https://doi.org/10.48550/arXiv.2508.05618)
47. DeepSeek-AI and collaborators. “[DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models](https://md.research.elamur.ai/free/en/papers/deepseek-ai-2025-deepseek-v3-2-pushing/)”. In: arXiv preprint arXiv:2512.02556 (2025). DeepSeek-V3.2 thinking-enhanced reasoning model technical report; accessed 2026-01-29. url: https://arxiv.org/abs/2512.02556. [↗](https://arxiv.org/abs/2512.02556)
48. An Yang, Anfeng Li, Baosong Yang, et al. [Qwen3 Technical Report](https://md.research.elamur.ai/free/en/papers/yang-2025-qwen3-technical-report/). 2025. doi: 10.48550/ARXIV.2505.09388. arXiv: 2505.09388. url: https://doi.org/10.48550/arXiv.2505.09388. [↗](https://doi.org/10.48550/arXiv.2505.09388)
49. AntAngelMed Team / Ant Group Health AI. AntAngelMed: Open-Source Medical Language Model. [https://github.com/MedAIBase/AntAngelMed](https://github.com/MedAIBase/AntAngelMed). Open-source medical reasoning model with MoE architecture and HealthBench-leading performance; accessed 2026-01-29. 2026. [↗](https://github.com/MedAIBase/AntAngelMed)
50. Yubin Kim, Hyewon Jeong, Shan Chen, et al. “Medical Hallucinations in Foundation Models and Their Impact on Healthcare”. In: CoRR abs/2503.05777 (2025). doi: 10.48550/ARXIV.2503.05777. arXiv: 2503.05777. url: https://doi.org/10.48550/arXiv.2503.05777. — [doi:10.48550/arxiv.2503.05777](https://doi.org/10.48550/arxiv.2503.05777) [↗](https://doi.org/10.48550/arXiv.2503.05777)
51. Itay Manes, Naama Ronn, David Cohen, Ran Ilan Ber, Zehavi Horowitz-Kugler, and Gabriel Stanovsky. “K-QA: A Real-World Medical Q&A Benchmark”. In: Proceedings of the 23rd Workshop on Biomedical Natural Language Processing, BioNLP@ACL 2024, Bangkok, Thailand, August 16, 2024. Ed. by Dina Demner-Fushma… [find](https://scholar.google.com/scholar?q=Itay+Manes%2C+Naama+Ronn%2C+David+Cohen%2C+Ran+Ilan+Ber%2C+Zehavi+Horowitz%2DKugler%2C+and+Gabriel+Stanovsky%2E+%E2%80%9CK%2DQA%3A+A+Real%2DWorld+Medical+Q%26A+Benchmark%E2%80%9D%2E+In%3A+Proceedings+of+the+23rd+Workshop+on+Biomedical+Natural)
52. Yuhui Li, Fangyun Wei, Chao Zhang, and Hongyang Zhang. “EAGLE-3: Scaling up Inference Acceleration of Large Language Models via Training-Time Test”. In: CoRR abs/2503.01840 (2025). doi: 10.48550/ARXIV.2503.01840. arXiv: 2503.01840. url: https://doi.org/10.48550/arXiv.2503.01840. — [doi:10.48550/arxiv.2503.01840](https://doi.org/10.48550/arxiv.2503.01840) [↗](https://doi.org/10.48550/arXiv.2503.01840)
53. Zihan Qiu, Zekun Wang, Bo Zheng, et al. “Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free”. In: CoRR abs/2505.06708 (2025). doi: 10.48550/ARXIV.2505.06708. arXiv: 2505.06708. url: https://doi.org/10.48550/arXiv.2505.06708. — [doi:10.48550/arxiv.2505.06708](https://doi.org/10.48550/arxiv.2505.06708) [↗](https://doi.org/10.48550/arXiv.2505.06708)
54. Zihao Zheng, Xiuping Cui, Size Zheng, et al. “MoQa: Rethinking MoE Quantization with Multi-stage Data-model Distribution Awareness”. In: CoRR abs/2503.21135 (2025). doi: 10.48550/ARXIV.2503.21135. arXiv: 2503.21135. url: https://doi.org/10.48550/arXiv.2503.21135. — [doi:10.48550/arxiv.2503.21135](https://doi.org/10.48550/arxiv.2503.21135) [↗](https://doi.org/10.48550/arXiv.2503.21135)
55. Wenhua Cheng, Weiwei Zhang, Haihao Shen, et al. “Optimize Weight Rounding via Signed Gradient Descent for the Quantization of LLMs”. In: Findings of the Association for Computational Linguistics: EMNLP 2024, Miami, Florida, USA, November 12-16, 2024. Ed. by Yaser Al-Onaizan, Mohit Bansal, and Yun-Nu… [find](https://scholar.google.com/scholar?q=Wenhua+Cheng%2C+Weiwei+Zhang%2C+Haihao+Shen%2C+et+al%2E+%E2%80%9COptimize+Weight+Rounding+via+Signed+Gradient+Descent+for+the+Quantization+of+LLMs%E2%80%9D%2E+In%3A+Findings+of+the+Association+for+Computational+Linguistics%3A+EMNL)
56. Elias Frantar, Saleh Ashkboos, Torsten Hoefler, and Dan Alistarh. “GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers”. In: CoRR abs/2210.17323 (2022). doi: 10.48550/ARXIV.2210.17323. arXiv: 2210.17323. url: https://doi.org/10.48550/arXiv.2210.17323. [↗](https://doi.org/10.48550/arXiv.2210.17323)
57. Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, et al. “[Training Verifiers to Solve Math Word Problems](https://md.research.elamur.ai/free/en/papers/cobbe-2021-training-verifiers-solve-math-word/)”. In: CoRR abs/2110.14168 (2021). arXiv: 2110.14168. url: https://arxiv.org/abs/2110.14168. [↗](https://arxiv.org/abs/2110.14168)
58. Mark Chen, Jerry Tworek, Heewoo Jun, et al. “[Evaluating Large Language Models Trained on Code](https://md.research.elamur.ai/free/en/papers/chen-2021-evaluating-large-language-models-trained/)”. In: CoRR abs/2107.03374 (2021). arXiv: 2107.03374. url: https://arxiv.org/abs/2107.03374. [↗](https://arxiv.org/abs/2107.03374)
59. Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, et al. “[Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://md.research.elamur.ai/free/en/papers/zheng-2023-judging-llm-judge-mt-bench/)”. In: Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023. Ed. by Al…
60. Shenggui Li, Yikai Zhu, Chao Wang, et al. SpecForge: Train speculative decoding models effortlessly. [https://github.com/sgl-project/specforge](https://github.com/sgl-project/specforge). 2025. [↗](https://github.com/sgl-project/specforge)
61. Lianmin Zheng, Liangsheng Yin, Zhiqiang Xie, et al. “[SGLang: Efficient Execution of Structured Language Model Programs](https://md.research.elamur.ai/free/en/papers/zheng-2023-sglang-efficient-execution-structured-language/)”. In: Advances in Neural Information Processing Systems 38: Annual Conference on Neural Information Processing Systems 2024, NeurIPS 2024, Vancouver, BC, Canada, December 10 - 15, 2…

*Updated 31 October 2026*