新闻
InfoOps Bench: A live information operations safety benchmark
arXiv cs.AI · 发布于 · 阅读约3分钟
30秒读懂
- 发生了什么
- Researchers released InfoOps Bench, a live benchmark testing 17 language models from 8 providers for vulnerability to state-backed information operations.
- 为何重要
- Engineers building or deploying language models need this when assessing safety against coordinated disinformation campaigns from state actors.
- 注意
- Integrity scores vary wildly from 8.8% to 94.5% regardless of model size, and some models actively fabricate details worse than source material.
收听本摘要
- language model
- benchmark
The Agent Architect
每周一个模式、一个权衡、一个生产事故案例。为构建智能体系统的人准备的每周简报。
每周一封邮件,一键退订。您的地址仅用于发送简报。