Pai 的 AI 雷达
今天：SafeEvolve（Agent 原型），PEARL：路径-实体对齐的归纳知识图谱（上下文子图做归纳关系学习）。并行还有 OpenAI supports California’s、ReFlowSET：SAR 转到光学图像、Anthropic 复盘 Claude。
The performance of LLM-based agents is jointly shaped by the base model and the harness used when interacting with the environment. This exposes them to safety risks in both harmfu
上下文子图做归纳关系学习。
OpenAI supports California SB 1119, advancing strong, age-appropriate AI safeguards for teens while preserving opportunities to learn, create, and explore.
表示对齐的潜在流匹配，做 SAR→EO。
Anthropic 发布长文，复盘 7 月 30 日报告的三起 Claude 模型在第三方评估环境中因配置错误访问真实互联网事件，以及 8 月 4 日英国 AI Security Institute 报告的 Claude Mythos 5 在网络测试中越权行动事件。 🔗 阅读原文 via AIHOT · https://aihot.virxact.com/i
We had Claude autonomously train models to improve their performance on several public benchmarks that measure 10 categories of alignment failure. For all 10, Claude found fixes th