A language model designed for working with scientific literature has received recognition from one of the most authoritative journals in the science world.
A year has passed since DeepSeek demonstrated that powerful models can be created without billion-dollar budgets – and the industry hasn't been the same since.
Lab
How to Teach AI to Discover New Things Right on the Dance Floor: Training Neural Networks During Testing
Computer Science
Researchers have taught a language model to find optimal solutions in science not through preliminary preparation, but by learning right in the process of working on a specific task.
AI: Events
Claude Taught to Write CUDA Kernels and Train Open Models
Technical context • Development
Anthropic has enhanced Claude's capabilities in handling low-level code and transferring knowledge to other models through its new «Extended Thinking» feature.
We explore the architectural solutions developers of Chinese open-source models are choosing and why decoder-based approaches continue to dominate the ecosystem.
AI: Events
Trinity Large: What's Inside and Why Arcee Released Three Versions of the Same Model
Technical context • Products
We dive into how Trinity Large from Arcee AI works as a new language model with a sparse architecture and three checkpoints to choose from.
AI: Events
Open Coding Agents: AI Code Assistants That Work With Any Repository
Technical context • Development
The Allen Institute for AI has unveiled Open Coding Agents – open-source models for autonomous coding that adapt to a project's structure.
AMD has unveiled ReasonLite-0.6B, a compact language model focusing on logical reasoning, trained using a majority voting strategy and a staged approach.
The compact GLM-4.7-Flash model is now available as an open-source solution, aiming to balance performance with the feasibility of running it on standard hardware.