Long-running models can solve hard tasks, but they expose safety risks short evals miss
polynoamial · x · 2026-07-21
Long-running models can solve hard open-ended tasks, but their persistence can also surface safety risks that short-horizon evaluations miss.
The post shares lessons from studying such a model and says the findings are shaping the team’s approach to:
- evaluations
- alignment
- monitoring
- user control
Related event: OpenAI Model Escapes Sandbox and Breaches Hugging Face(322 posts)→
More from Research
- Clinical System One v0.2: top clinical AI drops from 97% to 65% on harder cases — MaziyarPanahi · 2026-10-27
- IdeaAnchor mines 14K papers to teach LLMs literature-grounded ideation — armancohan · 2026-10-26
- MIT's Ziming Liu: AutoResearch doesn't generalize — blind training tricks cause 'forking' disasters — ZimingLiu11 · 2026-10-26
- Stanford Data-Centric AI Workshop reopens site, extends poster deadline — jure · 2026-10-26
- 'Vibe verification': ~60% of AI-written formal specs prove the wrong property — basedjensen · 2026-10-26
- Ito: a browser that detects browsing intent and auto-organizes tabs into Flows — andrewkuznet · 2026-10-26