Frequently updated list of dLLM (Diffusion Large Language Models) papers, models, and other resources

DeepSpec: a full-stack codebase for training and evaluating speculative decoding algorithms

A compiler, optimizer and executor for financial expressions and factors

TurboQuant: Near-optimal KV cache quantization for LLM inference (3-bit keys, 2-bit values) with Triton kernels + vLLM integration

A kernel library written in tilelang

An Asynchronous Reinforcement Learning Engine for Omni-Modal Post-Training at Scale

An agent-managed museum exhibit, built in Rust with Gajae-Code / LazyCodex — developed and maintained with no human intervention.

PTX ISA 9.1 documentation converted to searchable markdown. Includes Claude Code skill for CUDA development.

IndexCache: Accelerating Sparse Attention via Cross-Layer Index Reuse

A lightweight inference engine supporting speculative speculative decoding (SSD).

SGLang Omni: High-Performance Multi-Stage Pipeline Framework for Omni Models

Use PEFT or Full-parameter to CPT/SFT/DPO/GRPO 600+ LLMs (Qwen3.6, DeepSeek-V4, GLM-5.1, InternLM3, Llama4, ...) and 300+ MLLMs (Qwen3-VL, Qwen3-Omni, InternVL3.5, Ovis2.5, GLM4.5v, Gemma4, Llava, …

Autonomous GPU Kernel Generation & Optimization via Deep Agents

CUDA Agent: Large-Scale Agentic RL for High-Performance CUDA Kernel Generation

DFlash: Block Diffusion for Flash Speculative Decoding

Nsight Python is a Python kernel profiling interface based on NVIDIA Nsight Tools

Our first fully AI generated deep learning system

High Performance LLM Inference Operator Library

Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models

Open-source unified multimodal model

[ICLR 2025] DuoAttention: Efficient Long-Context LLM Inference with Retrieval and Streaming Heads

Accelerating Large-Scale Reasoning Model Inference with Sparse Self-Speculative Decoding

Accelerating MoE with IO and Tile-aware Optimizations

SLA: Beyond Sparsity in Diffusion Transformers via Fine-Tunable Sparse–Linear Attention

Tile-Based Runtime for Ultra-Low-Latency LLM Inference