Combining Double Soft-Min Critics with Adaptive KL Thresholding for Alignmet

H100-Optimized Goal-Conditioned Contrastive Self-RL for ARC Puzzles

Physics-based language model: O(n log n) wave field attention with linear-wave content routing. V4.1 achieves PPL 543 on WikiText-2.

The best-benchmarked open-source AI memory system. And it's free.

qqr is an RL training framework for open-ended agents.

James' cookbook of evaluations and finetuning experiments

AuON( Alternative Unit-norm momentum-updates by Normal- ized nonlinear scaling), a linear-time optimizer that achieves remarkable perfor- mance at linear time without producing semi-orthogonal matr…

Quick illustration of how one can easily read books together with LLMs. It's great and I highly recommend it.

The official repository for AdaMuon

Bootstrapping ARC

This Repo Contains Script To Fine Tune Open Source Models Using Unsloth by using UI with simple click and progress

Materials for ConceptARC paper

VeOmni: Scaling Any Modality Model Training with Model-Centric Distributed Recipe Zoo

Classify the morphologies of distant galaxies with cnn-vit

[CVPR2025] Breaking the Low-Rank Dilemma of Linear Attention

Leo optimizer, variation of Muon that runs faster

GLU Attention provide nearly cost-free performance boost for transformers with a simple mechanism that applies Gated Linear Unit to the values in Attention.

Awesome resources on normalizing flows.

codes for R-Zero: Self-Evolving Reasoning LLM from Zero Data (https://www.arxiv.org/pdf/2508.05004)

Stella Nera is the first Maddness accelerator achieving 15x higher area efficiency (GMAC/s/mm^2) and 25x higher energy efficiency (TMAC/s/W) than direct MatMul accelerators in the same technology

Unison file synchronizer

Continuous Thought Machines, because thought takes time and reasoning is a process.

American Sign Language Recognition System that translates Signs into their respective alphabets in real time

🤗 LeRobot: Making AI for Robotics more accessible with end-to-end learning

Lecture notes on the RL series provided by Stanford.

This is the homepage of a new book entitled "Mathematical Foundations of Reinforcement Learning."