arXiv · cs.AI · Humza Nusrat, Omar Nusrat · 4 小时前模型评分40
Claude Opus 5.5 用权重空间编程审计临床决策网络
Weight-space programming for auditing clinical decision networks
导读
Claude Opus 5.5 仅凭临床协议和网络权重，在无工具、无数据、无代码执行条件下盲审 18 个放疗计划筛选网络，在 16 个有缺陷网络中准确指出出错规则，并平均改动 3 个参数完成修复。研究还发现，针对协议测试能检出全部缺陷，黑盒测试则更快找到混淆因素和阈值偏移，但该方法尚不能证明网络完全正确。
分类置信度
中 · 待人工复核
- 只有单一媒体来源，需要人工复核。
来源级别：媒体与个人
应用领域：医疗与生命科学
事件类型：研究发布
证据摘录
- “The purpose of this study was to determine whether such a model, given only a clinical protocol and a network's weights, can audit and repair the network.”
- “Claude Opus 5.5 audited each network blind, without tools, data or code execution, and every claim and repair was tested on about 420,000 requests.”
- “Networks with 433 to 4,801 parameters were trained to screen head-and-neck planning requests against four rules.”
来源：arXiv · cs.AI · arxiv.org