跳到正文
原文
Rohan Paul· @rohanpaul_ai · X·· 12 天前AI 评分56
AI 导读

Overmind 推出面向智能体的训练平台,可记录生产环境中的智能体行为、生成训练数据,并微调和部署开放权重模型。客户可下载权重、重新训练或回滚,平台代码以 AGPL-3.0 开源。

正文

A 9B model beating a frontier flagship 4x on biomedical extraction is something.

@OvermindLab launched a platform that turns a company's own agent data into small specialist models the company owns.

It records what an agent does in production, turns those records into training data, then fine-tunes and serves an open-weight model.

Customers can download the weights, retrain them or roll back, and the platform code is open source under AGPL-3.0.

check out the below charts for its performance data. e.g. in one case across more than 4K questions on more than 100 contracts, Overmind's tuned 9B model found the right clause with 91.6% F1 against GPT-5.6 Luna's 82.0%.

引用Overmind@OvermindLab
Introducing Overmind. The training platform for agents. Overmind trained models across biomedical, legal, and engineering tasks beating the frontier and achieving: • 7x better accuracy • 20x cheaper usage • 28x less hallucinations Train your own: https://github.com/overmind-core/overmind
在 X 查看被引用的帖子

来源:Rohan Paul · x.com