The Xingchen Large-Scale Model Platform is an AI large-scale model collection platform launched by China Telecom, covering multimodal domains such as semantics, vision, and speech. It includes models with varying parameter levels, supports long text processing, and possesses multilingual speech recognition and multi-task capabilities...
Fast GraphRAG is a high-performance knowledge graph retrieval framework from Microsoft, designed to provide interpretable and high-precision agent-driven retrieval workflows. It combines Retrieval Augmentation (RAG) technology with knowledge graphs to improve the performance of large-scale...
Stagehand is a simple and extensible AI web browsing framework, the successor to Playwright. It provides three simple APIs: act, extract, and observe, and supports natural language-driven web page operations. Stagehand offers...
VideoTuna is a code library that integrates multiple AI video generation models, supporting text-to-video, image-to-video, and text-to-image conversions. VideoTuna provides comprehensive video processing capabilities including pre-training, continuous training, post-training alignment, and fine-tuning...
ComfyUI-MochiEdit is an open-source video editing tool based on ComfyUI and Genmo Mochi. It achieves video editing by converting video to noise and resampling, supporting partial editing and video-to-video conversion. Users can adjust...
HiCo is a hierarchical, controllable layout-to-image generation model developed by 360 AI Research Institute, based on a diffusion model. HiCo utilizes a multi-branch structure design to achieve precise control over object positioning and text description. HiCo's key features include...
Ferret-UI 2 is a multimodal, large-scale language model from Apple, used for understanding and interacting with mobile user interfaces. Ferret-UI 2 can recognize and understand UI elements on various mobile device screens, execute complex user commands,...
MMBench-Video is a novel long-video, multi-question answering benchmark jointly developed by Zhejiang University, Shanghai Artificial Intelligence Laboratory, Shanghai Jiao Tong University, and the Chinese University of Hong Kong. MMBench-Video can comprehensively evaluate large-scale visual...
MiniMates is a high-performance, lightweight digital human-driving algorithm with both voice and facial expression modes, capable of real-time operation on ordinary computers. Based on an advanced technical architecture, it delivers an exceptionally fast experience and personalized customization...
MotionCLR is an attention-based model for generating and editing human motion. It can generate actions based on text prompts and supports interactive editing by users, such as emphasizing, weakening, replacing, erasing, and style transfer. MotionCLR...
DocMind, developed by Sima Yue, is a large-scale intelligent document model based on the Transformer architecture. It integrates deep learning, NLP, and CV technologies to handle the complex structure and visual information of rich text documents, improving information extraction accuracy. DocM...
DistilQwen2 is a lightweight language model optimized from the Qwen2 large model using knowledge distillation techniques, which improves computational efficiency and reduces deployment costs. DistilQwen2 leverages deep model analysis and enhanced instruction data diversity...
Hunyuan3D-1.0 is a 3D generative model launched by Tencent. It supports both text and image input and generates high-quality 3D assets. The model employs a two-stage approach, first using a multi-view diffusion model to generate multi-view RGB...
Hunyuan-Large is a large-scale hybrid expert (MoE) model launched by Tencent, boasting 389 billion total parameters and 52 billion activation parameters, making it the largest open-source MoE model in terms of parameter size. Based on the Transformer architecture, it supports...
Cofounder is an open-source, full-stack AI development agent that helps developers automatically generate complete applications, including backend, frontend, database, and stateful web applications, based on a single prompt. Cofounder is based on generative AI technology...
MagicTailor is a new framework specifically designed for controllable personalization of components, enabling precise control over T2I models during the personalization process. MagicTailor is based on two key technologies: Dynamic Mask Degradation (DM-Deg) and Dual-Stream Balancing (...
Claude Computer Use is a new feature introduced by Anthropic for the Claude 3.5 Sonnet and Claude 3.5 Haiku models, enabling the Claude AI model to perform human-like computer operations, such as viewing the screen, moving light, etc.
Docling is an open-source document parsing and conversion tool that efficiently parses documents in various formats (including PDF, DOCX, PPTX, images, and HTML) and exports them to Markdown or JSON formats. Docling supports advanced PDF processing...
Cerebellum is a browser-based AI assistant built on Claude 3.5 Sonnet and Selenium WebDriver. It can understand task intent and perform automated web page tasks, such as data scraping and automated website testing. Cerebellum...
WebRL is a self-evolving online course reinforcement learning framework jointly launched by Tsinghua University and Zhipu AI. It trains high-performance network agents using open large language models (LLMs). WebRL dynamically generates tasks, supervises results, and provides rewards...
GameGen-X is a diffusion transformer model developed by researchers from institutions such as the Hong Kong University of Science and Technology and the University of Science and Technology of China. It's used to generate and interactively control open-world game videos. The model can simulate game engine functions, such as creating innovative characters...
OuteTTS is an open-source text-to-speech (TTS) project that generates speech based on pure language modeling. The OuteTTS project is based on the LLaMa architecture, using the Oute3-350M-DEV base model with 350 million parameters. OuteTTS features...
GTA (a benchmark for General Tool Agents) is a benchmark jointly launched by Shanghai Jiao Tong University and the Shanghai AI Lab to evaluate the ability of large language models (LLMs) to invoke tools in real-world scenarios.
VQAScore is an evaluation method jointly developed by CMU and Meta, which measures the quality of images generated from text prompts based on a visual question answering (VQA) model. VQAScore uses a computational model to evaluate the quality of images generated from text prompts, such as "Does this figure show {text}?"...