A platform for reproducible world model research and evaluation

Official PyTorch re-implementation of MiniT2I.

FlashRNN - Fast RNN Kernels with I/O Awareness

PyTorch implementation of RiT: Vanilla Diffusion Transformers Suffice in Representation Space

ARIS ⚔️ (Auto-Research-In-Sleep) — Lightweight Markdown-only skills for autonomous ML research: cross-model review loops, idea discovery, and experiment automation. No framework, no lock-in — works…

Multiplex Thinking: Reasoning via Token-wise Branch-and-Merge

PyTorch implementation of JiT https://arxiv.org/abs/2511.13720

Thinking with Videos from Open-Source Priors. We reproduce chain-of-frames visual reasoning by fine-tuning open-source video models. Give it a star 🌟 if you find it useful.

Accelerate TarFlow Sampling with GS-Jacobi Iteration

A unified media (Image, Video, Audio, Text) diffusion repository, for education and learning.

Code release for paper "Test-Time Training Done Right"

Video Generation, Physical Commonsense, Semantic Adherence, VideoCon-Physics

【CVPR 2025 Oral】Official Repo for Paper "AnyEdit: Mastering Unified High-Quality Image Editing for Any Idea"

A Pocket-Sized MLLM for Ultra-Efficient Image and Video Understanding on Your Phone

Official implementation of Inductive Moment Matching

[CVPR 2025 Oral] Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Models

[CVPR 2025] Sparse Voxels Rasterization: Real-time High-fidelity Radiance Field Rendering

🐳 Efficient Triton implementations for "Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention"

Lora traing script for Lightricks LTX-video

Official implementation of Continuous 3D Perception Model with Persistent State

Witness the aha moment of VLM with less than $3.

Simple RL training for reasoning

Code for NeurIPS 2024 paper - The GAN is dead; long live the GAN! A Modern Baseline GAN - by Huang et al.

Unofficial implementation of "Simplifying, Stabilizing & Scaling Continuous-Time Consistency Models" for MNIST

A framework that calibrates object properties through differentiable simulations of robot-object interactions.

OpenCoconut implements a latent reasoning paradigm where we generate thoughts before decoding.