Safe RLHF: Constrained Value Alignment via Safe Reinforcement Learning from Human Feedback

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

Native Multimodal Models are World Learners

Allegro is a powerful text-to-video model that generates high-quality videos up to 6 seconds at 15 FPS and 720p resolution from simple text input.

(NeurIPS 2024 Oral 🔥) Improved Distribution Matching Distillation for Fast Image Synthesis

(CVPR 2025) From Slow Bidirectional to Fast Autoregressive Video Diffusion Models

SEED-Voken: A Series of Powerful Visual Tokenizers

TextGen: Implementation of Text Generation models, include LLaMA, BLOOM, GPT2, BART, T5, SongNet and so on. 文本生成模型，实现了包括LLaMA，ChatGLM，BLOOM，GPT2，Seq2Seq，BART，T5，UDA等模型的训练和预测，开箱即用。

Unitree Go2, Unitree G1 support for Nvidia Isaac Lab (Isaac Gym / Isaac Sim)

minimal-cost for training 0.5B R1-Zero

[CVPR 2025 Highlight🔥] Identity-Preserving Text-to-Video Generation by Frequency Decomposition

UniWorld: High-Resolution Semantic Encoders for Unified Visual Understanding and Generation

High quality training free inpaint for every stable diffusion model. Supports ComfyUI

Implementation of MagViT2 Tokenizer in Pytorch

Deep Contextual Video Compression

⚡ Flash Diffusion ⚡: Accelerating Any Conditional Diffusion Model for Few Steps Image Generation (AAAI 2025 Oral)

USP: Unified (a.k.a. Hybrid, 2D) Sequence Parallel Attention for Long Context Transformers Model Training and Inference

A PyTorch implementation of MAGE: MAsked Generative Encoder to Unify Representation Learning and Image Synthesis

An MBTI Exploration of Large Language Models

[NeurIPS 2025 Spotlight] A Unified Tokenizer for Visual Generation and Understanding

MineWorld: A Real-time interactive world model on Minecraft

DiffusionNFT: Online Diffusion Reinforcement with Forward Process

NeurIPS 2024

Official inference code and LongText-Bench benchmark for our paper X-Omni (https://arxiv.org/pdf/2507.22058).

Benchmarking Legal Knowledge of Large Language Models

[ICCV 2025] VideoVAE+: Large Motion Video Autoencoding with Cross-modal Video VAE

GPT-ImgEval: Evaluating GPT-4o’s state-of-the-art image generation capabilities

[ICCV 2025] SimVQ: Addressing Representation Collapse in Vector Quantized Models with One Linear Layer

MoH: Multi-Head Attention as Mixture-of-Head Attention