arXiv cs.CL· Sina Daneshgar, Raya Namazi, Amirali Khosravi, Pouria Rouzrokh, Bardia Khosravi·· 17 小时前AI 评分38
EdgeDeID：无需真实患者数据训练的 0.6B 临床去标识化模型
EdgeDeID: A 0.6B Clinical De-identifier Trained Without Real Patient Data
AI 导读
EdgeDeID 是一个 0.6B 参数、完全本地运行的临床文本去标识化模型，训练全程不使用真实患者数据，由三个 LLM 教师生成合成病历并统一标注。在 100 条人工核验的合成域内基准上达到 0.994 micro-F1，99% 的病历无遗漏标识符，量化后可在 CPU 上以 615 MB 运行。
正文 · 原文
Abstract:De-identifying clinical notes with a hosted language model discloses the information the step is meant to protect. We present EdgeDeID, a 0.6B-parameter decoder-only de-identifier that runs locally and is trained without real patient text: three LLM teachers write synthetic notes, a single labeler re-annotates them into one convention, and the model learns to emit a typed XML schema of the identifiers it finds. On a 100-note synthetic in-domain benchmark with manually verified annotations it reaches 0.994 micro-F1 and leaves no missed identifier in 99% of notes, against 0.857 and 12% for the strongest of eight off-the-shelf systems. Quantized, it runs on CPU in 615 MB. Of three additions to supervised training, only a verifier adapter helps: it separates complete from incomplete outputs of other systems at 0.94 AUROC, whereas reinforcement learning with a recall-weighted reward and a rule-based post-processing layer leave in-domain results unchanged. On MIMIC ICU nursing notes F1 falls to 0.48. The errors follow two properties our synthetic notes lack, telegraphic style and low identifier density, and three hand-written in-context examples recover 0.08 F1 without annotated target data.
| Comments: | 18 pages, 2 figures. Under review at ACL Rolling Review (ARR). Code: this https URL ; models and data: this https URL |
| Subjects: | Computation and Language (cs.CL); Cryptography and Security (cs.CR); Machine Learning (cs.LG) |
| ACM classes: | I.2.7; J.3 |
| Cite as: | arXiv:2610.33839 [cs.CL] |
| (or arXiv:2610.33839v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.33839 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Sina Daneshgar [view email]
[v1]
Sat, 24 Oct 2026 18:20:52 UTC (154 KB)
来源：arXiv cs.CL · arxiv.org