Fireworks f1 is a composite AI model from Fireworks, designed for complex inference tasks. It achieves performance and reliability exceeding that of a single model by fusing multiple open models at the inference layer. The f1 model allows developers to...
AgileGen is a generative software development framework that enhances the software creation process through collaboration between humans and AI. AgileGen comprises two core components: end-user decision-making and the AgileGen intelligent agent. The framework is designed using the Gherkin programming language...
OmniSearch is a multimodal search enhancement and generation framework developed by Alibaba's Tongyi Lab, featuring adaptive planning capabilities. OmniSearch can dynamically decompose complex problems and adjust its search strategy based on search results and the problem context...
Perplexica is an open-source, AI-driven search engine and an open-source alternative to Perplexity AI. It uses machine learning algorithms and natural language processing to understand user queries and provide accurate answers. Perplexica supports various search methods...
Memoripy is a Python library that provides context-aware memory management for AI applications. Memoripy supports both short-term and long-term memory storage and is compatible with the OpenAI and Ollama APIs. Core functionalities include memory retrieval, concept extraction, ...
AnimateAnything is a unified and controllable video generation technology developed by researchers from Zhejiang University and Beijing University of Aeronautics and Astronautics. AnimateAnything can precisely manipulate video, including controlling camera trajectory, text prompts, and user gesture annotations...
RAG-Diffusion is a region-aware text-to-image generation method developed by a team at Nanjing University. Based on two stages—hard region binding and soft region thinning—it achieves precise control and detail optimization of each region in an image. RAG-Diffusion...
FitDiT is a high-fidelity virtual try-on technology jointly developed by Tencent and Fudan University. Based on Diffusion Transformers (DiT), it focuses on high-resolution features to enhance the presentation of clothing details. FitDiT extracts clothing textures...
Documind is an open-source AI document processing tool that can extract structured data from PDF documents. Documind can convert PDFs to images, extract information using the OpenAI API, and format output according to user-defined patterns...
MARS (Make vAriance Reduction Shine) is an innovative optimization framework launched by ByteDance to improve the efficiency of training large models. MARS integrates preconditioning gradient methods with variance reduction techniques, based on scaled stochastic recursion...
Verifier Engineering is a novel post-training paradigm jointly launched by the Chinese Academy of Sciences, Alibaba, and Xiaohongshu. It addresses the challenge of providing effective supervisory signals in basic model design. Verifier Engineering...
XiYan-SQL is a Natural Language to SQL (NL2SQL) framework launched by Alibaba. Based on a multi-generator integration strategy, it combines suggestive engineering and supervised fine-tuning to improve the quality of generated SQL queries. XiYan-SQL introduces M-Schema semi-structuring...
KuaiFormer is a Transformer-based retrieval framework developed by the Kuaishou technology team for use in large-scale content recommendation systems. It redefines the retrieval process, shifting from traditional score estimation tasks to a Transformer-driven approach...
EyeDiff is a text-to-image diffusion model that generates multimodal ophthalmic images based on natural language prompts, improving diagnostic accuracy for both common and rare eye diseases. The model is trained on multiple large-scale datasets and accurately captures key lesions...
BlueLM-V-3B is an algorithm and system co-design method jointly developed by vivo AI Lab and the MMLab of the Chinese University of Hong Kong. It supports the efficient deployment of multimodal large language models (MLLMs) to mobile devices. The model has a small size (2.7B)...
AtomThink is a multimodal mathematical reasoning framework jointly developed by researchers from Sun Yat-sen University, Hong Kong University of Science and Technology, Shanghai Jiao Tong University, the University of Hong Kong, and Huawei Noah's Ark Lab. The framework is based on the concept of building long chains of thought (CoT)...
DeepSeek-R1-Lite is a preview version of DeepSeek's inference model. Trained using reinforcement learning, DeepSeek-R1-Lite possesses long-chain reasoning capabilities and can display the reasoning process in real time. Its performance ranks highly in multiple benchmark tests...
Samsung Gauss2 is Samsung's second-generation multimodal generative AI model, designed to improve the performance and efficiency of Galaxy AI capabilities. Samsung Gauss2 can simultaneously process multiple data types, including text, code, and images.
AlphaQubit is an AI-based quantum error decoder launched by Google. It uses the Transformers deep learning architecture to identify and correct errors in quantum computing. Based on accurate error identification, AlphaQubit helps quantum computers...
CAD-MLLM is a computer-aided design (CAD) model generation system jointly developed by ShanghaiTech University, Transcengram, DeepSeek AI, and the University of Hong Kong. It generates models based on various user inputs (such as text descriptions, images, etc.).
LaTRO (Latent Reasoning Optimization) is an advanced framework that improves the performance of large language models (LLMs) on complex reasoning tasks. It is based on the analogy of the reasoning process to sampling from a latent distribution, using variational reasoning...
ReCapture is a video processing technology developed by Google and the National University of Singapore that can generate new videos with new camera trajectories from a single user-provided video. ReCapture uses a multi-view diffusion model or depth-based point cloud rendering...
MagicClay is a 3D modeling tool from Adobe that combines mesh and directed distance field (SDF) technology, allowing artists to sculpt specific parts of a 3D model based on text prompts while keeping other areas of the model unchanged. M...
StableV2V is an open-source video editing project launched by the University of Science and Technology of China. It enables precise editing and replacement of objects in videos based on inputs such as text, sketches, and images. The project uses a consistent shape editing paradigm and is based on three main components...