Tag

This paper introduces ReAct, an approach that integrates reasoning and acting in large language models (LLMs). ReAct enables LLMs to generate both reasoning traces and task-specific actions in an interleaved manner. This synergy allows reasoning to help induce, track, and update action plans, while actions interface with external sources like knowledge bases to gather more information, overcoming issues of hallucination and error propagation in prior methods.

SWE-agent is a system designed to empower language model (LM) agents to autonomously perform software engineering tasks. It features a custom agent-computer interface (ACI) that enhances the agent's ability to navigate repositories, create and edit code, and execute programs, achieving state-of-the-art results on the SWE-bench and HumanEvalFix benchmarks. [2, 5, 8]

LightRAG is an open-source framework designed for simple and fast Retrieval-Augmented Generation (RAG), integrating knowledge graphs, vector search, and efficient LLM-based processing to enhance question-answering over large document collections.

nanochat is a full-stack, minimal codebase for training, fine-tuning, evaluating, and deploying a ChatGPT-like large language model (LLM) from scratch on a single 8xH100 GPU node for under $100.

OpenCode is the open source AI coding agent built for the terminal. It features built-in agents (build, plan, general), LSP support, multi-provider LLM compatibility (75+ including Claude, GPT, Gemini, local), native TUI, multi-session, share links, and privacy-first design. No account needed, 35k+ stars.

A lightweight open-source platform for running, managing, and integrating large language models locally via a simple CLI and REST API.

Open-source framework that accelerates fine-tuning and full training of transformer LLMs by up to 30 × while cutting VRAM requirements by roughly 90 %, letting developers train custom models quickly on commodity GPUs.

Orchestrates and scales Python-based AI/ML workloads from laptop to thousands of GPUs — exposing task and actor primitives plus high-level libraries for training, hyperparameter tuning, serving, RL, and data processing. Designed for heterogeneous accelerators and production ML pipelines.

Curates step-by-step, hands-on tutorials for reimplementing technologies from scratch—covering everything from OSs and compilers to neural networks, LLMs, and vision systems—so learners learn by rebuilding real systems across languages.

Optimizes and deploys deep learning models for inference across CPU, GPU and Intel NPUs. Converts models from PyTorch/TensorFlow/ONNX, exposes C++/Python/NodeJS APIs, and integrates with Hugging Face and Optimum for LLM/GenAI workloads.

Provides unified model definitions and a single API for pretrained text, vision, audio, and multimodal models for both training and inference. Emphasizes cross-framework compatibility (PyTorch/TF/JAX), pipeline-based inference, and direct access to 1M+ Hub checkpoints.