DreamX-World: A General-Purpose Interactive World Model

[ICML2026] Auto-Regressive Long Video Generation via 2-Bit KV-Cache Quantization

verl Zero-Mismatch Dense/MoE HuggingFace Rollout

Muon in Int8 Precision Made Possible

Agent-friendly GPU profile-query CLI

A Claude Code skill for creating, compiling, reviewing, and polishing academic Beamer LaTeX presentations. Full lifecycle workflow with quality scoring, pedagogical review, TikZ audit, and more.

[CVPR 2026] CoMo: Learning Continuous Latent Motion from Internet Videos for Scalable Robot Learning

InfiniCCL is a unified, cross-platform collective communication library designed for heterogeneous accelerator environments.

high-performance inference and serving library for interactive autoregressive video and world models

Official Codebase for "DreamDojo: A Generalist Robot World Model from Large-Scale Human Videos" (ICML 2026)

Model compression toolkit engineered for enhanced usability, comprehensiveness, and efficiency.

Official Code of NAVA: Native Audio-Visual Alignment for Generation.

Official implementation of “Domino: Decoupling Causal Modeling from Autoregressive Drafting in Speculative Decoding”.

mKernel: fast multi-node, multi-GPU fused kernels

StreamDiffusion, Live Stream APP

CODA: Rewriting Transformer Blocks as GEMM-Epilogue Programs

Warp-as-History: Generalizable Camera-Controlled Video Generation from One Training Video

A Minimal and Elegant Framework & Tutorial for Real-Time Interactive World Models

Official implementation of Paper "System-Aware 4-Bit KV-Cache Quantization for Real-World LLM Serving"

CARE: Covariance-Aware and Rank-Enhanced Decomposition for Enabling Multi-Head Latent Attention

Triton kernels and PyTorch ops for Block Attention Residuals (AttnRes)

Hadamard transformation kernels written by cutedsl

Dataflow-Oriented Reinforcement Learning for (Multi-)Agentic LLMs