Models
- nvidia/Hymba-1.5B-Base · Hugging Face
- Cosmos Tokenizer: A suite of image and video neural tokenizersstar
- Nexusflow/Athene-V2-Agent · Hugging Face
Papers
- [2411.07975] JanusFlow: Harmonizing Autoregression and Rectified Flow for Unified Multimodal Understanding and Generation
- [2410.08020] Efficiently Learning at Test-Time: Active Fine-Tuning of LLMs
- [2411.10440] LLaVA-o1: Let Vision Language Models Reason Step-by-Step
- [2411.10433] M-VAR: Decoupled Scale-wise Autoregressive Modeling for High-Quality Image Generation
- [2411.13676] Hymba: A Hybrid-head Architecture for Small Language Models
- [2411.14402] Multimodal Autoregressive Pre-training of Large Vision Encodersstar
- [2411.14347] DINO-X: A Unified Vision Model for Open-World Object Detection and Understandingstar
- [2411.12155] Reinforcement Learning with Action Sequence for Data-Efficient Robot Learning
- [2411.14429] Revisiting the Integration of Convolution and Attention for Vision Backbone
Code
- GitHub - apple/ml-aim: This repository provides the code and model checkpoints for AIMv1 and AIMv2 research projects.
- GitHub - Lightricks/LTX-Video: Official repository for LTX-Video
- GitHub - rayleizhu/GLMix: [NeurIPS 2024] official code release for our paper “Revisiting the Integration of Convolution and Attention for Vision Backbone”.
Articles
- You could have designed state of the art positional encoding
- Extending the Context Length to 1M Tokens! | Qwen
Videos
- Tim Dettmers on Open-source AI, LMs, SWE Bench, Agents, Quantization, & Optimization - YouTube
- Speculations on Test-Time Scaling (o1) - YouTube
- Retrieval augmented generation; Extractive summarization - YouTube
- Learning at test time in LLMs - YouTube
- QA: Retrieval & Answer extraction - YouTube
- Flash Attention derived and coded from first principles with Triton (Python) - YouTube
- Guest Lecture 1: Or Patashnik - The Power of Attention Layers (KAIST CS492D, Fall 2024) - YouTube
Other
- CS 886: Recent Advances on Foundation Models
- Stanford Graph Learning Workshop 2024 | Stanford Engineering Data Science Applications
Tweets
: