Research applying AI agents to telecom infrastructure optimization and testing. Demonstrates agents in complex systems engineering.
Research advancing vision-language model capabilities for spatial grounding and object localization. Improves multimodal AI perception.
Anthropic releases detailed documentation on agent containment methods and two security incidents encountered. Transparency in agent safety practices.
Research paper on deploying mixture-of-experts models efficiently on mobile devices. Extends frontier of on-device AI capabilities.
Research paper presenting framework for agents to autonomously create, manage, and evaluate new skills. Advances state-of-art in agent self-improvement.
Security research identifying vulnerabilities in RLHF alignment processes that can be exploited to introduce misaligned behaviors. Critical safety finding.