---
title: "ICDAGENT — paper by Yin et al. (2026) — summary and ideas"
description: "They propose ICDAGENT, which pairs two fine-tuned open models. 28 references. Among them: 0."
authors: ["Ziyi Yin", "Yuanpu Cao", "Ting Wang", "Jinghui Chen", "Fenglong Ma"]
year: 2026
venue: "Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)"
source: "https://aclanthology.org/2026.acl-long.643/"
doi: "10.18653/v1/2026.acl-long.643"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/yin-2026-icdagent-empowering-agentic-large-language/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/yin-2026-icdagent-empowering-agentic-large-language/)

[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/36/) / ICDAGENT: Empowering Agentic Large Language Models for Explainable Medical Coding

# ICDAGENT: Empowering Agentic Large Language Models for Explainable Medical Coding

**Authors:** Ziyi Yin, [Yuanpu Cao](https://md.research.elamur.ai/free/en/authors/yuanpu-cao/), [Ting Wang](https://md.research.elamur.ai/free/en/authors/ting-wang/), [Jinghui Chen](https://md.research.elamur.ai/free/en/authors/jinghui-chen/) and [Fenglong Ma](https://md.research.elamur.ai/free/en/authors/fenglong-ma/) · **Year:** 2026 · **Venue:** [Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 2026](https://md.research.elamur.ai/free/en/venues/acl/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.acl-long.643/) · [PDF](https://aclanthology.org/2026.acl-long.643.pdf) · [DOI 10.18653/v1/2026.acl-long.643](https://doi.org/10.18653/v1/2026.acl-long.643) · [OpenAlex](https://openalex.org/W7166810402)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
The paper addresses explainable ICD coding: assigning International Classification of Diseases codes to clinical notes and justifying each code in free text. The authors argue that general-purpose commercial LLMs perform poorly here because they lack coding expertise and must navigate a huge label space of closely related codes. They propose ICDAGENT, which pairs two fine-tuned open models. A coding agent proposes codes with draft justifications. A critical agent reasons step by step to accept or reject each draft. To train the coding agent, the authors automatically built a rationale dataset from MIMIC-IV notes and official code descriptions. The critical agent first learns chain-of-thought from a proprietary teacher, then undergoes reinforcement learning with a reward that accounts for rationale quality. The two agents interact over several rounds. Across three datasets covering ICD-9 and ICD-10, ICDAGENT beats commercial LLMs and runs locally on two A100 GPUs. Local deployment lowers cost and avoids privacy risks.

## Key points
- **Coding agent:** HuatuoGPT-o1-8B is LoRA-fine-tuned on the new D_code dataset. The dataset was built from MIMIC-IV in two automated steps: an open-source LLM extracts a rationale for each gold code, then labels are aggregated.
- **Critical agent:** Qwen3-4B is LoRA-tuned on CoT trajectories distilled from Gemini. It is then optimized via RL with a rationale-aware reward. Rationales for incorrect codes are regenerated.
- **Baselines:** The commercial LLMs compared include GPT-4o, GPT-5-Chat, Claude-3.7/4-Sonnet, and Gemini-2.5-Pro/Flash. Evaluation uses MIMIC-III and MIMIC-IV, with predicted codes extracted by string matching on delimited tokens such as |E11.9|.
- **Comparison with CoRelation:** On micro-level metrics, where frequent codes carry most of the weight, ICDAGENT falls behind the classifier CoRelation. It outperforms CoRelation on macro-level metrics, indicating better handling of rare, long-tail codes.
- **Stated limitations:** Recall still lags behind state-of-the-art classifiers. Training also requires costly annotation, including commercial API use. The authors suggest future work on enriched code guidelines combined with RAG.

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

## Institutions

- [Pennsylvania State University](https://md.research.elamur.ai/free/en/universities/pennsylvania-state-university/) — University, State College, United States · 67 works · 5,865 citations

📚 Shelves (6): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🧬 Biochemistry, Genetics and Molecular Biology](https://md.research.elamur.ai/free/en/shelves/biochemistry-genetics-and-molecular-biology/) · 263 works · [🧬 Molecular Biology](https://md.research.elamur.ai/free/en/shelves/biochemistry-genetics-and-molecular-biology/molecular-biology/) · 260 works · [🏥 Health Professions](https://md.research.elamur.ai/free/en/shelves/health-professions/) · 75 works · [🏥 Health Information Management](https://md.research.elamur.ai/free/en/shelves/health-professions/health-information-management/) · 56 works

## Cite

Cite as: Ziyi Yin, Yuanpu Cao, Ting Wang, Jinghui Chen and Fenglong Ma. 2026. ICDAGENT: Empowering Agentic Large Language Models for Explainable Medical Coding. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

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

```bibtex
@inproceedings{yin2026icdagent,
  title = {{ICDAGENT: Empowering Agentic Large Language Models for Explainable Medical Coding}},
  author = {Ziyi Yin, Yuanpu Cao, Ting Wang, Jinghui Chen and Fenglong Ma},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.643},
  url = {https://doi.org/10.18653/v1/2026.acl-long.643}
}
```

- [← previous I2E: From Image Pixels to Actionable Interactive Environments for Text-Guided Image Editing](https://md.research.elamur.ai/free/en/papers/yu-2026-i2e-image-pixels-actionable-interactive/)
- [next → IceBreaker for Conversational Agents: Breaking the First-Message Barrier with Personalized Starters](https://md.research.elamur.ai/free/en/papers/zheng-2026-icebreaker-conversational-agents-breaking-first/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/36/)

## References

28 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Anthropic. 2025. Introducing claude 4. [https://www.anthropic.com/news/claude-4](https://www.anthropic.com/news/claude-4). Accessed: April 18, 2026. [↗](https://www.anthropic.com/news/claude-4)
2. Pengfei Cao, Yubo Chen, Kang Liu, Jun Zhao, Shengping Liu, and Weifeng Chong. 2020. Hypercore: Hyperbolic and co-graph representation for automatic ICD coding. In ACL, pages 3105–3114. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Pengfei+Cao%2C+Yubo+Chen%2C+Kang+Liu%2C+Jun+Zhao%2C+Shengping+Liu%2C+and+Weifeng+Chong%2E+2020%2E+Hypercore%3A+Hyperbolic+and+co%2Dgraph+representation+for+automatic+ICD+coding%2E+In+ACL%2C+pages+3105%E2%80%933114%2E+Association+for)
3. Junying Chen, Zhenyang Cai, Ke Ji, Xidong Wang, Wanlong Liu, Rongsheng Wang, Jianye Hou, and Benyou Wang. 2024. [Huatuogpt-o1, towards medical complex reasoning with llms](https://md.research.elamur.ai/free/en/papers/chen-2024-huatuogpt-o1-towards-medical-complex/). CoRR, abs/2412.18925.
4. Hua Cheng, Rana Jafari, April Russell, Russell Klopfer, Edmond Lu, Benjamin Striner, and Matthew Gormley. 2023. MDACE: MIMIC documents annotated with code evidence. In ACL (1), pages 7534–7550. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Hua+Cheng%2C+Rana+Jafari%2C+April+Russell%2C+Russell+Klopfer%2C+Edmond+Lu%2C+Benjamin+Striner%2C+and+Matthew+Gormley%2E+2023%2E+MDACE%3A+MIMIC+documents+annotated+with+code+evidence%2E+In+ACL+%281%29%2C+pages+7534%E2%80%937550%2E+Associ)
5. Gheorghe Comanici, Eric Bieber, Mike Schaekermann, Ice Pasupat, Noveen Sachdeva, Inderjit Dhillon, Marcel Blistein, Ori Ram, Dan Zhang, Evan Rosen, et al. 2025. [Gemini 2.5: Pushing the frontier with advanced reasoning, multimodality, long context, and next generation agentic capabilities](https://md.research.elamur.ai/free/en/papers/comanici-2025-gemini-2-5-pushing-frontier/). arXiv prep…
6. DeepSeek-AI, Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, Xiaokang Zhang, Xingkai Yu, Yu Wu, Z. F. Wu, Zhibin Gou, Zhihong Shao, Zhuoshu Li, Ziyi Gao, Aixin Liu, Bing Xue, Bingxuan Wang, Bochao Wu, Bei Feng, Chengda Lu, Chengg… — [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://md.research.elamur.ai/free/en/papers/deepseek-ai-2025-deepseek-r1-incentivizing-reasoning-capability/)
7. James C. Douglas, Yidong Gan, Ben Hachey, and Jonathan K. Kummerfeld. 2025. Less is more: Explainable and efficient ICD code prediction with clinical entities. In ACL (1), pages 30835–30847. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=James+C%2E+Douglas%2C+Yidong+Gan%2C+Ben+Hachey%2C+and+Jonathan+K%2E+Kummerfeld%2E+2025%2E+Less+is+more%3A+Explainable+and+efficient+ICD+code+prediction+with+clinical+entities%2E+In+ACL+%281%29%2C+pages+30835%E2%80%9330847%2E+Associati)
8. Joakim Edin, Maria Maistro, Lars Maaløe, Lasse Borgholt, Jakob D. Havtorn, and Tuukka Ruotsalo. 2024. An unsupervised approach to achieve supervised-level explainability in healthcare records. In EMNLP, pages 4869–4890. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Joakim+Edin%2C+Maria+Maistro%2C+Lars+Maal%C3%B8e%2C+Lasse+Borgholt%2C+Jakob+D%2E+Havtorn%2C+and+Tuukka+Ruotsalo%2E+2024%2E+An+unsupervised+approach+to+achieve+supervised%2Dlevel+explainability+in+healthcare+records%2E+In+EMNL)
9. Matús Falis, Aryo Pradipta Gema, Hang Dong, Luke Daines, Siddharth Basetti, Michael Holder, Rose S. Penfold, Alexandra Birch, and Beatrice Alex. 2024. Can GPT-3.5 generate and code discharge summaries? J. Am. Medical Informatics Assoc., 31(10):2284–2293. [find](https://scholar.google.com/scholar?q=Mat%C3%BAs+Falis%2C+Aryo+Pradipta+Gema%2C+Hang+Dong%2C+Luke+Daines%2C+Siddharth+Basetti%2C+Michael+Holder%2C+Rose+S%2E+Penfold%2C+Alexandra+Birch%2C+and+Beatrice+Alex%2E+2024%2E+Can+GPT%2D3%2E5+generate+and+code+discharge+summaries)
10. Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2021. [Lora: Low-rank adaptation of large language models](https://md.research.elamur.ai/free/en/papers/hu-2021-lora-low-rank-adaptation-large/). arXiv preprint arXiv:2106.09685.
11. Chao-Wei Huang, Shang-Chi Tsai, and Yun-Nung Chen. 2022. PLM-ICD: automatic ICD coding with pre-trained language models. CoRR, abs/2207.05289. — [arXiv:2207.05289](https://arxiv.org/abs/2207.05289) [find](https://scholar.google.com/scholar?q=Chao%2DWei+Huang%2C+Shang%2DChi+Tsai%2C+and+Yun%2DNung+Chen%2E+2022%2E+PLM%2DICD%3A+automatic+ICD+coding+with+pre%2Dtrained+language+models%2E+CoRR%2C+abs%2F2207%2E05289%2E+%E2%80%94+%5BarXiv%3A2207%2E05289%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2207%2E05289%29)
12. Fei Li and Hong Yu. 2020. ICD coding from clinical text using multi-filter residual convolutional neural network. In AAAI, pages 8180–8187. AAAI Press. [find](https://scholar.google.com/scholar?q=Fei+Li+and+Hong+Yu%2E+2020%2E+ICD+coding+from+clinical+text+using+multi%2Dfilter+residual+convolutional+neural+network%2E+In+AAAI%2C+pages+8180%E2%80%938187%2E+AAAI+Press%2E)
13. Rumeng Li, Xun Wang, and Hong Yu. 2024. Exploring LLM multi-agents for ICD coding. CoRR, abs/2406.15363. — [arXiv:2406.15363](https://arxiv.org/abs/2406.15363) [find](https://scholar.google.com/scholar?q=Rumeng+Li%2C+Xun+Wang%2C+and+Hong+Yu%2E+2024%2E+Exploring+LLM+multi%2Dagents+for+ICD+coding%2E+CoRR%2C+abs%2F2406%2E15363%2E+%E2%80%94+%5BarXiv%3A2406%2E15363%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2406%2E15363%29)
14. Leibo Liu, Óscar Pérez Concha, Anthony N. Nguyen, Vicki Bennett, and Louisa Jorm. 2022. Hierarchical label-wise attention transformer model for explainable ICD coding. J. Biomed. Informatics, 133:104161. [find](https://scholar.google.com/scholar?q=Leibo+Liu%2C+%C3%93scar+P%C3%A9rez+Concha%2C+Anthony+N%2E+Nguyen%2C+Vicki+Bennett%2C+and+Louisa+Jorm%2E+2022%2E+Hierarchical+label%2Dwise+attention+transformer+model+for+explainable+ICD+coding%2E+J%2E+Biomed%2E+Informatics%2C+133%3A1041)
15. Junyu Luo, Xiaochen Wang, Jiaqi Wang, Aofei Chang, Yaqing Wang, and Fenglong Ma. 2024. Corelation: Boosting automatic ICD coding through contextualized code relation learning. In LREC/COLING, pages 3997–4007. ELRA and ICCL. [find](https://scholar.google.com/scholar?q=Junyu+Luo%2C+Xiaochen+Wang%2C+Jiaqi+Wang%2C+Aofei+Chang%2C+Yaqing+Wang%2C+and+Fenglong+Ma%2E+2024%2E+Corelation%3A+Boosting+automatic+ICD+coding+through+contextualized+code+relation+learning%2E+In+LREC%2FCOLING%2C+pages+39)
16. George Michalopoulos, Michal Malyska, Nicola Sahar, Alexander Wong, and Helen H. Chen. 2022. Icdbig-bird: A contextual embedding model for ICD code classification. In BioNLP@ACL, pages 330–336. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=George+Michalopoulos%2C+Michal+Malyska%2C+Nicola+Sahar%2C+Alexander+Wong%2C+and+Helen+H%2E+Chen%2E+2022%2E+Icdbig%2Dbird%3A+A+contextual+embedding+model+for+ICD+code+classification%2E+In+BioNLP%40ACL%2C+pages+330%E2%80%93336%2E+Associ)
17. Andreas Motzfeldt, Joakim Edin, Casper L Christensen, Christian Hardmeier, Lars Maaløe, and Anna Rogers. 2025. Code like humans: A multi-agent solution for medical coding. arXiv preprint arXiv:2509.05378. [find](https://scholar.google.com/scholar?q=Andreas+Motzfeldt%2C+Joakim+Edin%2C+Casper+L+Christensen%2C+Christian+Hardmeier%2C+Lars+Maal%C3%B8e%2C+and+Anna+Rogers%2E+2025%2E+Code+like+humans%3A+A+multi%2Dagent+solution+for+medical+coding%2E+arXiv+preprint+arXiv%3A2509%2E05)
18. James Mullenbach, Sarah Wiegreffe, Jon Duke, Jimeng Sun, and Jacob Eisenstein. 2018. [Explainable prediction of medical codes from clinical text](https://md.research.elamur.ai/free/en/papers/mullenbach-2018-explainable-prediction-medical-codes-clinical/). In NAACL-HLT, pages 1101–1111. Association for Computational Linguistics.
19. Thanh-Tung Nguyen, Viktor Schlegel, Abhinav Ramesh Kashyap, and Stefan Winkler. 2023a. A two-stage decoder for efficient ICD coding. In ACL (Findings), pages 4658–4665. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Thanh%2DTung+Nguyen%2C+Viktor+Schlegel%2C+Abhinav+Ramesh+Kashyap%2C+and+Stefan+Winkler%2E+2023a%2E+A+two%2Dstage+decoder+for+efficient+ICD+coding%2E+In+ACL+%28Findings%29%2C+pages+4658%E2%80%934665%2E+Association+for+Computational+L)
20. John Schulman, Philipp Moritz, Sergey Levine, Michael I. Jordan, and Pieter Abbeel. 2016. High-dimensional continuous control using generalized advantage estimation. In ICLR (Poster). [find](https://scholar.google.com/scholar?q=John+Schulman%2C+Philipp+Moritz%2C+Sergey+Levine%2C+Michael+I%2E+Jordan%2C+and+Pieter+Abbeel%2E+2016%2E+High%2Ddimensional+continuous+control+using+generalized+advantage+estimation%2E+In+ICLR+%28Poster%29%2E)
21. John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. 2017. Proximal policy optimization algorithms. CoRR, abs/1707.06347. [find](https://scholar.google.com/scholar?q=John+Schulman%2C+Filip+Wolski%2C+Prafulla+Dhariwal%2C+Alec+Radford%2C+and+Oleg+Klimov%2E+2017%2E+Proximal+policy+optimization+algorithms%2E+CoRR%2C+abs%2F1707%2E06347%2E)
22. Guangming Sheng, Chi Zhang, Zilingfeng Ye, Xibin Wu, Wang Zhang, Ru Zhang, Yanghua Peng, Haibin Lin, and Chuan Wu. 2025. [Hybridflow: A flexible and efficient rlhf framework](https://md.research.elamur.ai/free/en/papers/sheng-2024-hybridflow-flexible-efficient-rlhf-framework/). In Proceedings of the Twentieth European Conference on Computer Systems, pages 1279–1297.
23. Thanh Vu, Dat Quoc Nguyen, and Anthony N. Nguyen. 2020. A label attention model for ICD coding from clinical text. In IJCAI, pages 3335–3341. ijcai.org. [find](https://scholar.google.com/scholar?q=Thanh+Vu%2C+Dat+Quoc+Nguyen%2C+and+Anthony+N%2E+Nguyen%2E+2020%2E+A+label+attention+model+for+ICD+coding+from+clinical+text%2E+In+IJCAI%2C+pages+3335%E2%80%933341%2E+ijcai%2Eorg%2E)
24. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, Jian Yang, Jianhong Tu, Jianwei Zhang, Jian Yang, Jiaxi Yang, Jingren Zhou, Jun… — [Qwen3 Technical Report](https://md.research.elamur.ai/free/en/papers/yang-2025-qwen3-technical-report/)
25. Zheng Yuan, Chuanqi Tan, and Songfang Huang. 2022. Code synonyms do matter: Multiple synonyms matching network for automatic ICD coding. In ACL (2), pages 808–814. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Zheng+Yuan%2C+Chuanqi+Tan%2C+and+Songfang+Huang%2E+2022%2E+Code+synonyms+do+matter%3A+Multiple+synonyms+matching+network+for+automatic+ICD+coding%2E+In+ACL+%282%29%2C+pages+808%E2%80%93814%2E+Association+for+Computational+Lingui)
26. Tong Zhou, Pengfei Cao, Yubo Chen, Kang Liu, Jun Zhao, Kun Niu, Weifeng Chong, and Shengping Liu. 2021. Automatic ICD coding via interactive shared representation networks with self-distillation mechanism. In ACL/IJCNLP (1), pages 5948–5957. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Tong+Zhou%2C+Pengfei+Cao%2C+Yubo+Chen%2C+Kang+Liu%2C+Jun+Zhao%2C+Kun+Niu%2C+Weifeng+Chong%2C+and+Shengping+Liu%2E+2021%2E+Automatic+ICD+coding+via+interactive+shared+representation+networks+with+self%2Ddistillation+mecha)
27. Thanh-Tung Nguyen, Viktor Schlegel, Abhinav Ramesh Kashyap, Stefan Winkler, Shao-Syuan Huang, Jie-Jyun Liu, and Chih-Jen Lin. 2023b. Mimic-iv-icd: A new benchmark for extreme multilabel classification. CoRR, abs/2304.13998. — [arXiv:2304.13998](https://arxiv.org/abs/2304.13998) [find](https://scholar.google.com/scholar?q=Thanh%2DTung+Nguyen%2C+Viktor+Schlegel%2C+Abhinav+Ramesh+Kashyap%2C+Stefan+Winkler%2C+Shao%2DSyuan+Huang%2C+Jie%2DJyun+Liu%2C+and+Chih%2DJen+Lin%2E+2023b%2E+Mimic%2Div%2Dicd%3A+A+new+benchmark+for+extreme+multilabel+classification)
28. OpenAI. 2025. Gpt-5 system card. [https://cdn.openai.com/gpt-5-system-card.pdf](https://cdn.openai.com/gpt-5-system-card.pdf). Technical report describing GPT-5 safety evaluations and capabilities. Accessed: April 18, 2026. [↗](https://cdn.openai.com/gpt-5-system-card.pdf)

*Updated 2 November 2026*