Intellectual hub of the topic
AI: Events
Technical context • Infrastructure
AMD has unveiled a method for partitioning a single GPU into isolated domains to run different models simultaneously – with no compromise on security or performance.
AI21 Labs has developed a method enabling language models to accurately extract text from documents in right-to-left languages.
Researchers at AI21 Labs have devised a method to reduce online reinforcement learning costs by sending some data to «take a nap» until better times.
MiniMax has discussed its approach to fine-tuning language models that do more than just answer questions – they execute complex tasks by interacting with tools.
Over 90% of Salesforce's engineers now write code in Cursor, which has noticeably sped up development and improved code quality.
AI: Events
Technical context • Development
Mistral AI engineers shared how they tracked down a memory leak in the popular vLLM language model serving system and what challenges they faced.
Overworld has released Waypoint-1, a video generation model that runs locally and responds to controls in real-time during the content creation process.
AMD has unveiled TileLang – a tool that simplifies writing optimized GPU operators and lowers the barrier to entry for ROCm development.
The startup Overcut uses Azure to build secure agent systems that help companies automate work within complex development infrastructure.
Want to know about new
experiments first?
Subscribe to our Telegram channel — we share all the latest
and exciting updates from NeuraBooks.