No, we put in on Livebench and it’s not so good
RESEARCH
-
Open-source datasets remain bottleneck for AI model training
By
–
I think the recipe is here from papers and open-source work from Hugging Face, Allen AI, and many others, but open-source datasets are not quite there yet : (
-

SWE-1.5 fastest coding agent released by Cognition Cerebras
By
–
Today, @cognition released SWE-1.5 – the world’s fastest coding agent, powered by Cerebras. SWE-1.5 achieves frontier-level coding ability, comparable to Sonnet 4.5 and surpassing GPT-5. Cerebras and Cognition engineers worked hand in hand over the past few weeks, training a
-

Linearizer Collapses Diffusion Model Training to Single Step
By
–
Who Said Neural Networks Aren’t Linear?? In this paper, authors are able to collapse the training of diffusion models down to only 1 step by introducing Linearizer, which sandwiches a linear matrix A between two invertible neural networks!
-
Elsevier Expands Geographical Pricing for Open Access
By
–
Elsevier has announced it is expanding its Geographical Pricing for Open Access to support greater participation in global research. https://
bit.ly/4qxmkkT
#openaccess #Elsevier #research -
RL Progress: From Human-Driven Intuition to Direct Machine Search
By
–
why this breaks everything: RL progress has been bottlenecked by human intuition. researchers have insights, try variations, publish. it takes years to go from Q-learning to DQN to PPO. now you just let the machine search directly. millions of variants in weeks instead of
-
DeepMind AI Develops Self-Improving Learning Algorithms
By
–
TL;DR for normal people: imagine you're teaching a robot to learn. humans spent decades figuring out the "best ways" to teach machines (called learning algorithms). deepmind built an AI that invents its own teaching methods. and they work better than ours. why it matters: →
-

DeepMind AI discovers new reinforcement learning algorithms from scratch
By
–
deepmind just published something wild they built an AI that discovers its own reinforcement learning algorithms. not hyperparameter tuning. not tweaking existing methods. discovering ENTIRELY NEW learning rules from scratch. and the algorithms it found were better
-
2D Gaussian Splatting: Real-Time Neural Rendering and Reconstruction
By
–
🚀2D Gaussian Splatting: Real-Time, Geometry-Aware Radiance Field Reconstruction
— Satya Mallick (@LearnOpenCV) 29 octobre 2025
In this week’s deep dive, we unpack how 2D Gaussian Splatting (2DGS) redefines the future of real-time neural rendering and reconstruction. By collapsing volumetric 3D Gaussians into surface-aligned… pic.twitter.com/FBuCopT64x2D Gaussian Splatting: Real-Time, Geometry-Aware Radiance Field Reconstruction In this week’s deep dive, we unpack how 2D Gaussian Splatting (2DGS) redefines the future of real-time neural rendering and reconstruction. By collapsing volumetric 3D Gaussians into surface-aligned
-

US Should Explore Alternative Computing Approaches Beyond Datacenters
By
–
Besides being wildly unsustainable the datacenter boom assumes one way of building AI. The US should back efforts to explore new frontiers in computing.