Lil2J

turbo-fieldfare. Gemma 4 26B-A4B inference in ~2 GB of RAM on any M-series MacBook

seta. 💻 SETA: Scaling Environments for Terminal Agents

grok-build. SpaceXAI's coding agent harness and TUI. Fullscreen, mouse interactive, extensible.

oh-my-pi. ⌥ AI Coding agent for the terminal — hash-anchored edits, optimized tool harness, LSP, Python, browser, subagents, and more

labs-molt. Python

SWE-smith. [NeurIPS 2025 D&B Spotlight] Scaling Data for SWE-agents

ResearchClawBench. 🦞 ResearchClawBench: Evaluating AI Agents for Automated Research from Re-Discovery to New-Discovery

openscience. The open-source AI workbench for scientific research

Automodel. 🚀 Pytorch Distributed native training library for LLMs/VLMs with OOTB Hugging Face support

MegaDLMs. GPU-optimized framework for training diffusion language models at any scale. The backend of Quokka, Super Data Learners, and OpenMoE 2 training.

orbit. Stable and Efficient Reinforcement Learning for Trillion-Parameter LLMs

speculators. A unified library for building, evaluating, and storing speculative decoding algorithms for LLM inference in vLLM

lucebox. LLM speculative inference server for consumer hardware & heterogeneous computing

RL_Envs_101. Building and Scaling RL environments in the age of LLMs

warp. Warp is an agentic development environment, born out of the terminal.

SWE-agent. SWE-agent takes a GitHub issue and tries to automatically fix it, using your LM of choice. It can also be employed for offensive cybersecurity or competitive coding challenges. [NeurIPS 2024]

Reagent. Agent-RRM: Exploring Reasoning Reward Model for Agents

Open-dLLM. Open diffusion language model for code generation — releasing pretraining, evaluation, inference, and checkpoints.

OpenMythos. A theoretical reconstruction of the Claude Mythos architecture, built from first principles using the available research literature.

minimax_search. MiniMax Search is an MCP (Model Context Protocol) server that provides web search and browsing capabilities.

SkillClaw. Let Skills Evolve Collectively with Agentic Evolver

hermes-agent. The agent that grows with you

SWE-Gym. Code for Paper: Training Software Engineering Agents and Verifiers with SWE-Gym [ICML 2025]

cuLA. CUDA kernels for linear attention variants, written in CuTe DSL and CUTLASS C++.

Trinity-RFT. Trinity-RFT is a general-purpose, flexible and scalable framework designed for reinforcement fine-tuning (RFT) of large language models (LLM).

polarquant-kv. LLM KV Cache compression - K+V dual compression, 73-99% VRAM savings, zero accuracy loss

ToolRL. Python

verl-tool. A version of verl to support diverse tool use [TMLR 2026]

A-mem. A-MEM: Agentic Memory for LLM Agents

CoMLRL. Open-Source Library for Fully Cooperative Multi-LLM Reinforcement Learning

MARTI. A Framework for LLM-based Multi-Agent Reinforced Training and Inference

oh-my-codex. OmX - Oh My codeX: Your codex is not alone. Add hooks, agent teams, HUDs, and so much more.

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

PostTrainBench. Measuring how well CLI agents like Claude Code or Codex CLI can post-train base LLMs on a single H100 GPU in 10 hours

pi. AI agent toolkit: unified LLM API, agent loop, TUI, coding agent CLI

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

supermemory. Memory and context engine + app that is extremely fast, scalable, and can be run fully locally. The Memory API for the AI era.

lsbot. Lean & Secure Bot

claw-eval. Claw-Eval is an evaluation harness for evaluating LLM as agents. All tasks verified by humans.

MetaClaw. 🦞 Just talk to your agent — it learns and EVOLVES 🧬.

Attention-Residuals.

clash-for-linux-install. 😼 优雅地使用基于 clash/mihomo 的代理环境

OpenShell. OpenShell is the safe, private runtime for autonomous AI agents.

NemoClaw. Run agents like Hermes, LangChain Deep Agents, and OpenClaw more securely inside NVIDIA OpenShell with managed inference

Tool-Star. 🔧Tool-Star: Empowering LLM-brained Multi-Tool Reasoner via Reinforcement Learning

MiroFlow. 🏆 Top-1 on 5+ benchmarks | Web UI | Supports MiroThinker, Claude, Kimi, OpenAI

CrossWOZ. A Large-Scale Chinese Cross-Domain Task-Oriented Dialogue Dataset

cerul. Cerul — where video becomes citable. Search video by meaning — across speech, visuals, and on-screen text.

CSL. [COLING 2022] CSL: A Large-scale Chinese Scientific Literature Dataset 中文科学文献数据集

autoresearch. AI agents running research on single-GPU nanochat training automatically

LightAgent. LightAgent: Lightweight Python framework for OpenAI-compatible agents with tools, memory, guardrails, tracing, lifecycle hooks, multi-agent collaboration, and workflows.

slime-agentic. A project implementing various agentic RL based on the Slime post-training framework

OpenClaw-RL. OpenClaw-RL: Train any agent simply by talking

Acontext. Agent Skills as a Memory Layer

CLUEDatasetSearch. 搜索所有中文NLP数据集，附常用英文NLP数据集

ml-engineering. Machine Learning Engineering Open Book

Self-Evolving-Agents.

skills. Give your agents the power of the Hugging Face ecosystem

nanoclaw. A lightweight alternative to OpenClaw that runs in containers for security. Connects to WhatsApp, Telegram, Slack, Discord, Gmail and other messaging apps,, has memory, scheduled jobs, and runs directly on Anthropic's Agents SDK

ljg-explain-concept. Claude Code skill: 8-dimensional concept anatomist. Deconstructs any concept into epiphany.

SDPO. Reinforcement Learning via Self-Distillation (SDPO)

ms-swift. 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, Phi4, ...) (AAAI 2025).

ruflo. 🌊 The leading agent meta-harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, RAG integration, and native Claude Code / Codex / Hermes and many more Integrated

openclaw. Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞

QeRL. [ICLR 2026]QeRL enables RL for 32B LLMs on a single H100 GPU.

Megatron-Bridge. Training library for Megatron-based models with bidirectional Hugging Face conversion capability

SkyRL. SkyRL: A Modular Full-stack RL Library for LLMs

ms-agent. MS-Agent: a lightweight framework to empower agentic execution of complex tasks

agentscope. Build and run agents you can see, understand and trust.

tvm-ffi. Open ABI and FFI for Machine Learning Systems

learn-claude-code. Bash is all you need - A nano claude code–like 「agent harness」, built from 0 to 1

Skill_Seekers. Convert documentation websites, GitHub repositories, and PDFs into Claude AI skills with automatic conflict detection

Fast-dLLM. Official implementation of "Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding"

hello-agents. 📚 《从零开始构建智能体》——从零开始的智能体原理与实践教程

AionUi. Free, local, open-source 24/7 Cowork app for OpenClaw, Hermes Agent, Claude Code, Codex, OpenCode, Gemini CLI and 20+ more CLI | Customize your assistants | Star if you like it!

hpc-ops. High Performance LLM Inference Operator Library

drzero. Dr. Zero Self-Evolving Search Agents without Training Data

CLUECorpus2020. Large-scale Pre-training Corpus for Chinese 100G 中文预训练语料

skillsbench. SkillsBench evaluates how well skills work and how effective agents are at using them.

ToRL. Python

eigent. Eigent: The Open Source Cowork Desktop - Local and Free Alternative to Claude Cowork and Codex

cc-switch. A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Grok Build & Hermes Agent. Only official website: ccswitch.io

opencode. The open source coding agent.

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

SimpleMem. SimpleMem: Efficient Lifelong Memory for LLM Agents — Text & Multimodal

awesome-LLM-driven-kernel-generation. Review automated kernel generation in the era of LLMs

AgentFlow. AgentFlow: In-the-Flow Agentic System Optimization

dflash. DFlash: Block Diffusion for Flash Speculative Decoding

MiroThinker. MiroThinker is a deep research agent optimized for complex research and prediction tasks. Our latest models, MiroThinker-1.7, achieves 74.0 and 75.3 on the BrowseComp and BrowseComp Zh, respectively.

wtian. Vue

VeOmni. VeOmni: Scaling Any Modality Model Training with Model-Centric Distributed Recipe Zoo

R1-Searcher. R1-searcher: Incentivizing the Search Capability in LLMs via Reinforcement Learning

OpenTinker. OpenTinker is an RL-as-a-Service infrastructure for foundation models

Search-R1. Search-R1: An Efficient, Scalable RL Training Framework for Reasoning & Search Engine Calling interleaved LLM based on veRL

mini-sglang. A compact implementation of SGLang, designed to demystify the complexities of modern LLM serving systems.

A-mem-sys. A-MEM: Agentic Memory for LLM Agents

Scripts-SGLang. Shell

dInfer. dInfer: An Efficient Inference Framework for Diffusion Language Models

LLaDA2.X. LLaDA2.0 is the diffusion language model series developed by InclusionAI team, Ant Group.

dllm. dLLM: Simple Diffusion Language Modeling

MemAgent. A MemAgent framework that can be extrapolated to 3.5M, along with a training framework for RL training of any agent workflow.

Awesome-Interleaving-Reasoning. Interleaving Reasoning: Next-Generation Reasoning Systems for AGI

FlexKV. Python

cutlass. CUDA Templates and Python DSLs for High-Performance Linear Algebra

vllm-omni. A framework for efficient model inference with omni-modality models

Super-Experts-Profilling. (ICLR 2026) Unveiling Super Experts in Mixture-of-Experts Large Language Models

flashinfer. FlashInfer: Kernel Library for LLM Serving

prime-rl. Agentic RL Training at Scale

ToolOrchestra. ToolOrchestra is an end-to-end RL training framework for orchestrating tools and agentic workflows.

auto-round. A SOTA quantization algorithm for high-accuracy low-bit LLM inference, seamlessly optimized for CPU/XPU/CUDA, with multi-datatype support and full compatibility with vLLM, SGLang, and Transformers.

miles. Miles is an enterprise-facing reinforcement learning framework for LLM and VLM post-training, forked from and co-evolving with slime.

reap. REAP: Router-weighted Expert Activation Pruning for SMoE compression

Awesome-ML-SYS-Tutorial. My learning notes for ML SYS.

InfinityStar. [NeurIPS 2025 Oral]Infinity⭐️: Unified Spacetime AutoRegressive Modeling for Visual Generation

tinker-cookbook. Post-training with Tinker

nano-PEARL. Draft-Target Disaggregation LLM Serving System via Parallel Speculative Decoding.

prism-research. Research prototype of PRISM — a cost-efficient multi-LLM serving system with flexible time- and space-based GPU sharing.

kvcached. Virtualized Elastic KV Cache for Dynamic GPU Sharing and Beyond

agent-lightning. The absolute trainer to light up AI agents.

nanochat. The best ChatGPT that $100 can buy.

tensorrtllm_backend. The Triton TensorRT-LLM Backend

Speech. A scalable generative AI framework built for researchers and developers working on Large Language Models, Multimodal, and Speech AI (Automatic Speech Recognition and Text-to-Speech)

TensorRT-LLM. TensorRT LLM provides users with an easy-to-use Python API to define Large Language Models (LLMs) and supports state-of-the-art optimizations to perform inference efficiently on NVIDIA GPUs. TensorRT LLM also contains components to create Python and C++ runtimes that orchestrate the inference execution in a performant way.

Model-Optimizer. A unified library of SOTA model optimization techniques like quantization, distillation, pruning, neural architecture search, speculative decoding, etc. It compresses deep learning models for downstream deployment frameworks like TensorRT-LLM, TensorRT, vLLM, etc. to optimize inference speed.

llm-compressor. Transformers-compatible library for applying various compression algorithms to LLMs for optimized deployment with vLLM

tilelang. Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels

ai-agents-for-beginners. 18 Lessons to Get Started Building AI Agents

aibrix. Cost-efficient and pluggable Infrastructure components for GenAI inference

LeetCUDA. LeetCUDA: Modern CUDA Learn Notes with PyTorch for Beginners, 200+ CUDA Kernels, Tensor Cores, HGEMM, FA-2 MMA.

flash-attention-minimal. Flash Attention in ~100 lines of CUDA (forward pass only)

ai-goofish-monitor. 基于 Playwright 和AI实现的闲鱼多任务实时/定时监控与智能分析系统，配备了功能完善的后台管理UI。帮助用户从闲鱼海量商品中，找到心仪产品。

Flash-RL. Implementation for FP8/INT8 Rollout for RL training without performence drop.

ultrascale-playbook-zh. UltraScale Playbook 中文版

pansou. PanSou是一款高性能的网盘资源搜索API服务，支持TG频道和插件搜索。系统设计以性能和可扩展性为核心，支持多频道多插件并发搜索、结果智能排序和网盘类型分类。docker集成前后端，一键启动，开箱即用。仅供学习研究，请勿以各种形式用于盈利目的。

LMCache. LMCache: Supercharge Your LLM with the Fastest KV Cache Layer

focalboard. Focalboard is an open source, self-hosted alternative to Trello, Notion, and Asana.

Megatron-LM. Ongoing research training transformer models at scale

slime. slime is an LLM post-training framework for RL Scaling.

SpecForge. Train speculative decoding models effortlessly and port them smoothly to SGLang serving.

Triton-distributed. Distributed Compiler based on Triton for Parallel Systems

nn-zero-to-hero. Neural Networks: Zero to Hero

crawl4ai. 🚀🤖 Crawl4AI: Open-source LLM Friendly Web Crawler & Scraper. Don't be shy, join here: https://discord.gg/jP8KfhDhyN

Metis-RISE. Metis-RISE: RL Incentivizes and SFT Enhances Multimodal Reasoning Model Learning

Mac-list. Mac软件清单、Mac使用技巧整理，正在不断完善中。努力做到最全。

gemlite. Fast low-bit matmul kernels in Triton

how-to-optim-algorithm-in-cuda. how to optimize some algorithm in cuda.

sgl-learning-materials. Materials for learning SGLang

clash-verge-rev. A modern GUI client based on Tauri, designed to run in Windows, macOS and Linux for tailored proxy experience

LLM-RL-Visualized. 🌟100+ 原创 LLM / RL 原理图📚，《大模型算法》作者巨献！💥（100+ LLM/RL Algorithm Maps ）

build-your-own-x. Master programming by recreating your favorite technologies from scratch.

ebooks. 收藏的一些经典的历史、政治、心理、哲学、数学、计算机方面电子书(约10万本）

ART. Agent Reinforcement Trainer: train multi-step agents for real-world tasks using GRPO. Give your agents on-the-job training. Reinforcement learning for Qwen3.6, GPT-OSS, Llama, and more!

OpenBB. Open Data Platform for analysts, quants and AI agents.

RL2. Python

Kimi-K2. Kimi K2 is the large language model series developed by Moonshot AI team

Triton-Puzzles. Puzzles for learning Triton

nanoVLM. The simplest, fastest repository for training/finetuning small-sized VLMs.

FlagGems. FlagGems is an operator library for large language models implemented in the Triton Language.

GenAI_Agents. 50+ tutorials and implementations for Generative AI Agent techniques, from basic conversational bots to complex multi-agent systems.

CS-Books. 🔥🔥超过1000本的计算机经典书籍、个人笔记资料以及本人在各平台发表文章中所涉及的资源等。书籍资源包括C/C++、Java、Python、Go语言、数据结构与算法、操作系统、后端架构、计算机系统知识、数据库、计算机网络、设计模式、前端、汇编以及校招社招各种面经~

prompt-eng-interactive-tutorial. Anthropic's Interactive Prompt Engineering Tutorial

12-factor-agents. What are the principles we can use to build LLM-powered software that is actually good enough to put in the hands of production customers?

DeepResearch. Tongyi Deep Research, the Leading Open-source Deep Research Agent

triton_tutorial. Tutorials for Triton, a language for writing gpu kernels

gemini-cli. An open-source AI agent that brings the power of Gemini directly into your terminal.

KsanaLLM. C++

FlagScale. FlagScale is a large model toolkit based on open-sourced projects.

ai-hedge-fund. An AI Hedge Fund Team

text-to-lora. Hypernetworks that adapt LLMs for specific benchmark tasks using only textual task description as the input

nano-vllm. Nano vLLM

ROLL. An Efficient and User-Friendly Scaling Library for Reinforcement Learning with Large Language Models

LUFFY. Official Repository of "Learning to Reason under Off-Policy Guidance"

Absolute-Zero-Reasoner. Official Repository of Absolute Zero Reasoner

Intuitor. [ICLR 2026] Learning to Reason without External Rewards

agent-distillation. Official Code Repository for the paper "Distilling LLM Agent into Small Models with Retrieval and Code Tools"

DeepEyes. Python

ParScale. Parallel Scaling Law for Language Model — Beyond Parameter and Inference Time Scaling

verl-agent. verl-agent is an extension of veRL, designed for training LLM/VLM agents via RL. verl-agent is also the official code for paper "Group-in-Group Policy Optimization for LLM Agent Training"

BaldEagle. 3x Faster Inference; Unofficial implementation of EAGLE Speculative Decoding

simpleRL-reason. Simple RL training for reasoning

picotron. Minimalistic 4D-parallelism distributed training framework for education purpose

Seed-Thinking-v1.5.

Semi-PD. A prefill & decode disaggregated LLM serving framework with shared GPU memory and fine-grained compute isolation.

APR. [COLM 2025] Code for Paper: Learning Adaptive Parallel Reasoning with Language Models

Tina. [ICLR 2026] Tina: Tiny Reasoning Models via LoRA

legalbenchrag. This is the repo for the LegalBench-RAG Paper: https://arxiv.org/abs/2408.10343.

TokenSwift. [ICML 2025] |TokenSwift: Lossless Acceleration of Ultra Long Sequence Generation

speculative_thinking. Python

specreason. PoC for "SpecReason: Fast and Accurate Inference-Time Compute via Speculative Reasoning" [NeurIPS '25]

SpeculativeDecodingPapers. 📰 Must-read papers and blogs on Speculative Decoding ⚡️

Spec-Bench. Spec-Bench: A Comprehensive Benchmark and Unified Evaluation Platform for Speculative Decoding (ACL 2024 Findings)

Jakiro. This repository is the official implementation of "Jakiro: Boosting Speculative Decoding with Decoupled Multi-Head via MoE" [ACL 2026 Main Accepted]

Medusa. Medusa: Simple Framework for Accelerating LLM Generation with Multiple Decoding Heads

HASS. Official Implementation of "Learning Harmonized Representations for Speculative Sampling" (HASS)

AReaL. The RL Bridge for LLM-based Agent Applications. Made Simple & Flexible.

EAGLE. Official Implementation of EAGLE-1 (ICML'24), EAGLE-2 (EMNLP'24), and EAGLE-3 (NeurIPS'25).

ParallelSpeculativeDecoding. [ICLR 2025] PEARL: Parallel Speculative Decoding with Adaptive Draft Length

self-speculative-decoding. Code associated with the paper **Draft & Verify: Lossless Large Language Model Acceleration via Self-Speculative Decoding**

dynamo. A Datacenter Scale Distributed Inference Serving Framework

FlashMLA. FlashMLA: Efficient Multi-head Latent Attention Kernels