InsightTok: Improving Text and Face Fidelity in Discrete Tokenization for Autoregressive Image Generation Paper β’ 2605.14333 β’ Published 8 days ago β’ 32
Running 93 Scaling FineWeb to 1000+ languages: Step 1: finding signal in 100s of evaluation tasks π 93 Evaluate multilingual models using FineTasks
Running on CPU Upgrade 236 The Synthetic Data Playbook: Generating Trillions of the Finest Tokens π 236 Explore synthetic data experiments on a virtual bookshelf
AnyFlow: Any-Step Video Diffusion Model with On-Policy Flow Map Distillation Paper β’ 2605.13724 β’ Published 9 days ago β’ 96
MHLA: Restoring Expressivity of Linear Attention via Token-Level Multi-Head Paper β’ 2601.07832 β’ Published Jan 12 β’ 53
Running Featured 1.35k FineWeb: decanting the web for the finest text data at scale π· 1.35k Explore and download the FineWeb webβtext dataset
Running 3.85k The Ultra-Scale Playbook π 3.85k The ultimate guide to training LLM on large GPU Clusters
Running on CPU Upgrade Featured 3.18k The Smol Training Playbook π 3.18k The secrets to building world-class LLMs