arXiv cs.CL· Davide Vitabile, Alexandro Buffa, Akshay Nambiar, Amril Nazir·· 11 小时前AI 评分55
MedPsy：面向端侧部署的 1.7B 与 4B 医疗语言模型
MedPsy: State-of-the-Art Small Medical Language Models for Efficient Edge Deployment
AI 导读
MedPsy 推出面向端侧部署的纯文本 1.7B 和 4B 医疗语言模型，训练配方结合合成医疗数据管线与 235B 医疗教师模型的思维链目标，并采用两阶段 SFT 加两阶段 RL 的后训练课程。
正文 · 原文
Abstract:Medical language models promise expert clinical support, but the strongest open-source LLMs remain too large for private edge-device deployment, while compact medical models still trail substantially on knowledge-intensive and real-world healthcare tasks. Reasoning models that fit edge memory budgets can produce long reasoning traces, increasing sequential decoding, end-to-end latency, energy consumption, and KV-cache memory pressure. We present MedPsy, a family of text-only 1.7B and 4B medical language models designed for edge deployment, targeting this accuracy-generation-efficiency trade-off. Our recipe combines a synthetic medical data pipeline over biology, medicine, and health seeds with chain-of-thought targets from a 235B medical-focused LLM teacher; a four-stage post-training curriculum comprising two SFT stages followed by two RL stages with hard-sample mining; and an edge deployment evaluation spanning multiple GGUF quantization variants and real-device inference on laptop and smartphone hardware. On seven closed-ended medical benchmarks, MedPsy-4B surpasses MedGemma-1.5-4B by 19.34 points and matches MedGemma-27B while being 6.75x smaller; MedPsy-1.7B outperforms MedGemma-1.5-4B by 11.42 points despite being less than half its size. On HealthBench-Hard, MedPsy-4B surpasses MedGemma-27B by 15.33 points, while MedPsy-1.7B surpasses it by 11.66 points despite being ~16x smaller. Beyond accuracy, MedPsy reduces average response length by 1.7x (1.7B) and 3.2x (4B) versus its Qwen3 backbones, while Q4_K_M quantization preserves the aggregate score within 1 point of BF16 with ~69% lower disk footprint. In real-device inference, MedPsy-4B's shorter generations translate into 75% lower median end-to-end latency and estimated energy per answer than its Qwen3 backbone at matched quantization, while peak RAM remains below 5 GB across our tested laptop and smartphone configurations.
| Comments: | Accepted at NeurIPS 2026 (main conference) |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.02011 [cs.CL] |
| (or arXiv:2611.02011v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.02011 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Amril Nurman Nazir [view email]
[v1]
Sun, 1 Nov 2026 15:52:32 UTC (992 KB)
来源：arXiv cs.CL · arxiv.org