Hi [First name],
Here are your weekly recs!
Upcoming events
|
How Discrete Diffusion Speeds Up LLMs
Friday, November 13th 10:00-11:00 AM PT
Subham Sahoo will walk through Uno, a project with IFM and Cerebras that uses discrete diffusion to generate several tokens at once. The team reports up to 3× faster generation while preserving the original model’s output distribution. Join us on Zoom to hear how it works and ask Subham your questions.
Sign Up Here
|
|
Trending papers
|
View all
|
Understanding AI from the Inside: The State of Mechanistic Interpretability for AI Safety
Neural networks are likely to be among the most consequential artefacts in human history. As these systems become more capable and autonomous, their p...
13,990 views
| |
Mistral Large 4
We introduce Mistral Large 4 (ML4), a 52B active, 1T total parameter multimodal mixture-of-experts model designed for reasoning and long-horizon agent...
19,686 views
|
|
Together Compiler: Continual Compilation for AI Inference
Today’s AI workloads are rapidly changing: new models are released more often, model architectures are increasingly diverse, and traffic distributions...
24,905 views
| |
You don't need long context
The study tests whether tasks with long interaction histories or large inputs can be handled using short active contexts. Across agent benchmarks, periodic s...
4,142 views
|
|
Stabilizing On-Policy Targeted Guidance in Agentic Reinforcement Learning
Naive on-policy RL is often bottlenecked by the capabilities of the base model --- insufficient coverage can result in an inability to learn from cert...
2,151 views
| |
A Survey on Continuous Diffusion and Flow Language Models
Continuous diffusion and flow language models generate text by transforming noise into continuous representations of tokens. They can refine many posi...
4,311 views
|
|
|