Urgent.News

What's breaking now, across thousands of outlets.

AI

EvoRiskBench: An Evolving Benchmark for Runtime Security Risks in Workspace Agents

Workspace agents combine large language models with execution harnesses to perform stateful, multi-step tasks that access or modify external resources. Existing benchmarks leave gaps in executable coverage of their runtime security risks, while evolving model capabilities, harnesses, tools, and threats motivate benchmark evolution. We introduce EvoRiskBench, an evolving benchmark organized around…

We haven't written up this one. arXiv cs.AI has the full story — the link below goes straight to it.

Read the original at arxiv.org →

More in AI

I built an AI-assisted I Ching app for Android — here's what I learned about digital divination

I'm a software PM from China who got into the Yijing (I Ching) last year. I went looking for a good app and found two extremes: walls of classical Chinese that bounce beginners instantly, or horoscope…

  • BU is a free AI-assisted I Ching app for Android, bilingual in English and Chinese.
  • Users find BU more as a journaling tool than traditional fortune-telling, driven by daily fortunes.

CodeSage: Code Intelligence for AI Agents, and Why I Rebuilt Its Models

Until this week, CodeSage ran its two models exactly as they ship on Hugging Face: Jina's code embedder and the ms-marco MiniLM reranker.

  • CodeSage rebuilt with new models and improved performance
  • Full index of php-src now generates in 373 seconds
  • Rebuilt models available on Hugging Face with CLI and MCP server

More from Friday 2 October →