跳到正文
𝕏 Rohan Paul(x.com)·· 2 小時前精選品質評分78
重點導讀

醫療 AI 助理要回答醫師的真實問題,也可能得查詢醫院病歷;這些工作需要不同能力。

  • Ant Group 發布了 560B 引數的醫療 MoE 模型 LingSante-3.1-flash,並表示它在 HealthBench Professional、HLE、BrowseComp 和 EHR-Complex 四項基準上超越 GPT-6 Astra、Claude Opus 5.5 和 Gemini 4 Argon。EHR-Complex 測試模型查詢醫院病歷資料庫的能力。
  • 模型權重已放在 Hugging Face,採用 MIT 授權。

AI 伴隨問答

自備 Key 深度分析

針對這篇報導有疑問?可直接點選快捷問題或向 AI 追問背景脈絡與技術細節:

推薦理由

LingSante-3.1-flash在四項醫療與通用基準上領先多個模型,呈現醫療推理、病歷查詢與工具使用能力的比較結果。

Rohan Paul
Rohan Paul@rohanpaul_ai
·2 小時前
𝕏

Ant Group released LingSante-3.1-flash, a 560B-parameter medical MoE model, and it tops GPT-6 Astra, Claude Opus 5.5 and Gemini 4 Argon on HealthBench Professional, HLE, BrowseComp and EHR-Complex.

HealthBench Professional cover answering real clinician requests that real doctors bring to chatbots. EHR-Complex tests an agent querying a hospital records database.

weights on Huggingface, with MIT-license.

Ant mixed medical and general data from pretraining onward, trained medical reasoning, tool use, search and clinician-style behavior as separate skills with their own RL rewards on physician- and verifier-checked tasks, then merged those specialists into one model through multi-teacher on-policy distillation.

推文來源 · X (Twitter)在 X 參與討論

讀者討論

理性交流 · 即時討論