AndroidLab is a framework for training and evaluating Android autonomous agents, integrating text and image modal manipulation environments, a unified action space, and reproducible benchmarks. AndroidLab supports large language models and multimodal models...
Recraft V3 is an AI text-to-image generation model developed by Recraft. It ranked first on Hugging Face's text-to-image leaderboard with an ELO score of 1172. The model features high-quality image generation and advanced...
Magentic-One is a general-purpose multi-agent system from Microsoft, designed to solve complex network and file tasks across various domains. The system is based on a multi-agent architecture, led by the Orchestrator agent, which coordinates WebSurfer, FileSurfer, and more...
X-Portrait 2 is a single-image video-driven technology developed by ByteDance's Intelligent Creation Team. It generates high-quality, cinematic videos based on a single still photo and a driving video clip. X-Portrait 2 preserves the original image's identity features and accurately...
Ichigo is an open-source, multimodal AI voice assistant that employs a mixed-modal model to process interleaved sequences of speech and text in real time. It directly quantizes speech into discrete tokens and uses a unified transformer architecture to simultaneously process speech and text...
CogSound, the latest sound effect model from Zhipu AI, adds captivating sound effects to silent videos. Based on GLM-4V's video understanding capabilities, CogSound accurately identifies and understands the semantics and emotions behind videos, enhancing the sound of silent videos...
AgentSquare is a framework developed by a team at Tsinghua University for automatically searching and optimizing agents for large language model (LLM) systems. Based on standardized module interface abstractions, it enables high-speed self-evolution and adaptive evolution of AI agents. The framework includes...
Fashion-VDM is a virtual try-on technology developed jointly by Google and the University of Washington, based on the Video Diffusion Model (VDM). It can generate high-quality virtual try-on images of a person wearing a given garment image and a video of the person...
Olly.bot is a personal AI assistant integrated into iMessage and SMS. Based on OpenAI's large models, it provides functions such as web search, document analysis, and image generation. Olly.bot requires no download or registration, does not collect user identity information, and protects...
AdaCache (Adaptive Caching) is an open-source technology from Meta that accelerates the AI video generation process. AdaCache's adaptive caching mechanism optimizes the allocation of computing resources, dynamically adjusting the computational load based on the complexity of different video content...
CogVideoX v1.5 is the latest open-source AI video generation model from Zhipu. The model includes two versions: CogVideoX v1.5-5B and CogVideoX v1.5-5B-I2V. The 5B series models support generating videos 5 to 10 seconds long, at 768P resolution, and with 16...
MiLoRA is a parameter-efficient fine-tuning method for large language models (LLMs) that reduces computational and memory costs by updating the secondary singularities of the weight matrix. The method is based on singular value decomposition (SVD) to divide the weight matrix into primary and secondary components...
OpenCoder is an open-source large-scale language model (LLM) developed by researchers from the University of Melbourne, Fudan University, and other universities in collaboration with Infinite Lightyear. It improves the performance of open-source LLMs to the level of proprietary models, advancing research in code AI...
Infinity-MM is a multimodal instruction dataset with tens of millions of samples, released by the Beijing Academy of Artificial Intelligence (BAAI). It contains 43 million samples and 10TB of data. The dataset has undergone quality filtering and deduplication to ensure high quality and diversity, and can improve open-source vision...
FabricDiffusion is a high-fidelity 3D clothing generation technology jointly developed by Google and Carnegie Mellon University. It can transfer the textures and prints of real-world 2D clothing images to 3D clothing models of arbitrary shapes with high quality.
HourVideo is a long-form video understanding benchmark dataset developed by Fei-Fei Li and Jia-Jun Wu's team at Stanford University. It contains 500 first-person perspective videos, ranging in length from 20 to 120 minutes, covering 77 daily activities, and can evaluate the performance of multimodal models...
App Intents is a new framework introduced by Apple that allows developers to integrate Siri and Apple Intelligence on platforms such as iOS and macOS, enabling application functionality and system experience (such as Siri, Spotlight search, Shortcuts, etc.).
DimensionX is a framework jointly developed by the Hong Kong University of Science and Technology, Tsinghua University, and BioNTech. It can generate highly realistic 3D and 4D scenes from a single image, achieving precise control over spatial and temporal dimensions based on video diffusion technology. ...
SeedEdit is a general-purpose image editing model developed by ByteDance's Doubao Big Model team. It's based on simple natural language commands for image editing, including retouching, clothing replacement, enhancement, style transfer, and adding or deleting elements in specified areas.
HK-O1aw is the world's first large-scale legal reasoning model based on the slow-thinking paradigm, developed through a collaboration between the AI for Reasoning team (HKAIR) under the Hong Kong Generative AI Research and Development Center (HKGAI) and the Alignment Team of Peking University (PKU).
AnimePro FLUX is an anime-style image generation model based on the Apache 2.0 license, specifically designed for generating high-quality 2D illustrations. AnimePro FLUX is a finely tuned version of the Flux.1 Shnell model, overcoming the licensing limitations of the DEV version...
VideoChat is an open-source, real-time digital human dialogue system that supports voice input and real-time conversation. Users can customize the digital human's appearance and voice, and voice cloning is possible without training. First-packet latency can be as low as 3 seconds, making it suitable for...
TableGPT2 is a novel large-scale multimodal model developed by Zhejiang University, designed for the integration and processing of tabular data. It is the first model to train structured data as an independent modality, directly understanding and manipulating data from databases, Excel files, etc.
GenXD is a 3D-4D co-generation framework jointly developed by the National University of Singapore and Microsoft. It can generate high-quality 3D and 4D scenes from any number of conditional images. The framework uses a data processing workflow to extract camera data from video...