Research & Education

CodeShell is a robust series of code large language models developed by PKU-KCL, featuring 7 billion parameters and trained on 500 billion tokens with an 8192 context window. It excels in code evaluation benchmarks like HumanEval and MBPP, outperforming similar-sized models such as CodeLlama and Starcoder. The project offers a comprehensive ecosystem including base models, chat models (with 4-bit quantization for reduced memory), and C++ versions for local deployment without a GPU. CodeShell also provides IDE plugins for VS Code and JetBrains, along with various demo options including Web-UI, CLI, and OpenAI API compatibility, making it a versatile solution for code generation, fill-in-the-middle tasks, and code-related Q&A.

Ficha offers an AI-powered solution for characterizing waste quality, designed to enhance sorting and collection performance for municipalities, collection operators, and landlords. The platform utilizes computer vision to analyze waste composition in various environments, providing granular data that helps improve waste management strategies. Key offerings include "Le Cocon" for collective habitats, which characterizes sorting quality per deposit, improves vertical habitat sorting, and automates communication. "La Camia" is tailored for collection operators, enabling detection of sorting errors at bin collection, mapping sorting quality across territories, and pre-emptively decluttering trucks. Ficha aims to provide actionable insights to optimize waste management and increase recycling rates.

CellCo, also known as The Cell Company, is a therapeutics company dedicated to engineering breakthrough medicines. They achieve this by integrating advanced molecular simulation techniques with artificial intelligence. The company's core focus is on deep technology development, aiming to innovate in the pharmaceutical and biotech sectors. Their approach involves using AI to model and engineer biological systems, which is crucial for the discovery and development of novel therapeutic solutions. CellCo is actively expanding its team and capabilities to further its mission of creating impactful medicines.

Moonlight is an AI-powered PDF reader designed to streamline the research process for academics and students. It integrates AI features directly into the document interface, allowing users to get instant explanations, summaries, and translations of research papers without needing to switch between applications. The tool provides context-aware AI chat, citation checking, and a library for organizing papers. Moonlight aims to lower the barrier to understanding complex research, offering features like auto-highlighting, markup tools, and scholar deep search to enhance productivity and comprehension. It is trusted by researchers globally for its ability to make reading papers easier and more efficient.

GPTSwarm is a sophisticated graph-based framework designed for building and managing LLM-based agents. It offers two primary high-level features: the ability to construct agents from graphs and to facilitate the customized, automatic self-organization of agent swarms, complete with self-improvement functionalities. The library includes components for environment definition, graph creation and execution, LLM backend interfacing, index-based memory, and optimization algorithms to enhance agent performance and swarm efficiency. GPTSwarm supports local LLM inference via LM Studio and provides clear instructions for quick setup and execution of agent swarms, making it a powerful tool for researchers and developers in agentic AI.

Youtu-Agent is a flexible, high-performance open-source framework designed for building, running, and evaluating autonomous agents. It excels in delivering powerful agent capabilities, such as data analysis, file processing, and deep research, primarily utilizing open-source models like DeepSeek-V3. The framework supports automated agent generation through Workflow and Meta-Agent modes, simplifying tool code, prompt, and configuration creation. Additionally, it features continuous experience learning via Training-Free GRPO for in-context optimization and a scalable Agent RL module for end-to-end reinforcement learning, addressing stability and scalability challenges. Youtu-Agent is optimized for low-cost deployment and offers practical use cases including literature review, personal file organization, and PPT generation.

SpatialVLA is a cutting-edge, spatial-enhanced vision-language-action model designed for robotics research and development. Trained on an extensive dataset of 1.1 Million real robot episodes, it aims to advance robot learning and performance in complex visual-language-action tasks. The model is built on a purely HuggingFace-based, concise code structure, ensuring efficient performance and easy deployment. It supports both pre-training from scratch using large datasets like OXE and RH20T, and fine-tuning with LoRA for real-world experiments with smaller datasets. SpatialVLA also provides evaluation code for benchmarks like SimplerEnv and offers a Model Zoo with various pre-trained and fine-tuned models.

Helmholtz AI is an application-driven artificial intelligence platform dedicated to advancing and democratizing AI capabilities for tackling frontier scientific questions and grand societal challenges within the Helmholtz Association. The platform provides essential resources such as AI consulting, offering expertise to leverage AI methods in research. It also grants low-barrier GPU access for accelerating innovative AI applications and secures funding through project calls to kick-off collaborations. Helmholtz AI actively promotes collaboration and ensures accessibility to critical AI resources and expertise, fostering an open community for AI development and implementation across various scientific fields.

NLTK (Natural Language Toolkit) is a leading open-source platform designed for building Python programs to work with human language data. It offers a comprehensive suite of text processing libraries for tasks such as classification, tokenization, stemming, tagging, parsing, and semantic reasoning. NLTK also provides easy-to-use interfaces to over 50 corpora and lexical resources, including WordNet, and includes wrappers for industrial-strength NLP libraries. The toolkit is accompanied by a hands-on guide that introduces programming fundamentals alongside computational linguistics topics, making it suitable for linguists, engineers, students, educators, researchers, and industry users. NLTK is available for Windows, macOS, and Linux, and is supported by an active discussion forum and a community-driven project.

Parakeet-tdt_ctc-1.1b is an AI speech recognition model available as a Hugging Face Space by NVIDIA. This tool allows users to upload or record audio and receive a transcript that includes precise timestamps for each segment of speech. It is designed to process audio and provide a detailed breakdown of when each part of the speech begins and ends, making it suitable for applications requiring accurate temporal alignment of text and audio. While the live website currently shows a build error, the intended functionality is to offer robust speech-to-text capabilities with a focus on detailed timing information.

Play with SD Models is a free AI tool hosted on Hugging Face Spaces, designed for users to experiment with various Stable Diffusion models. This platform facilitates hands-on experience with AI-powered image generation, making it suitable for educational and research purposes. While the space is currently paused, it aims to provide a straightforward interface for exploring the capabilities of different Stable Diffusion architectures. Users interested in understanding how these models work or in generating unique images can utilize this tool, once reactivated, to test prompts and observe outputs without needing extensive technical setup.

Chi SquareX provides comprehensive data analytics and statistical solutions, helping businesses and researchers transform raw data into actionable insights. The platform specializes in cutting-edge statistical analysis, machine learning solutions, and data visualization to uncover patterns and drive strategic decision-making. With a focus on delivering competitive advantages, Chi SquareX offers expert analytics services tailored to various needs. The tool aims to simplify complex data challenges, making advanced analytical techniques accessible for deeper understanding and informed strategies. It supports both individual researchers and corporations in their data-driven endeavors.

Mool.ai is an AI-powered platform designed to accelerate and enhance research for strategy and investment professionals. It delivers institutional-grade intelligence across all companies and markets, enabling users to discover opportunities and build conviction. The platform provides continuously updated fundamental views, forensic analysis, driver-based valuations, and catalyst-driven scenarios. Users can request on-demand analysis, including company research, sector views, peer benchmarks, and valuation analysis, which are produced instantly. Mool.ai also offers bespoke deliverables formatted to internal or client templates, such as equity reports, models, memos, and presentations. It covers mid, small, and micro-caps, emerging markets, and includes real-time global data across 200+ exchanges and 90+ countries, along with private company profiles in select geographies.

LG AI Research is dedicated to advancing artificial intelligence for a better life, focusing on its potential to transform industries, drive economic growth, and address global challenges. The organization explores various AI domains, including Superintelligence, EXAONE Language, Physical Intelligence, Bio Intelligence, Data Intelligence, and Materials Intelligence. Through its research, LG AI Research aims to empower individuals and organizations to leverage AI's capabilities. The initiative also features the EXAONE Showroom, highlighting practical applications and solutions derived from their cutting-edge research.

End-to-End Driving at Scale 2024 is a platform designed for participants in autonomous driving competitions. Users can access comprehensive details about the competition, including rules, datasets, and specific requirements. The tool facilitates the submission process for participants' models and provides real-time updates on leaderboards, allowing for performance tracking and comparison. Hosted on Hugging Face Spaces, it serves as a central hub for researchers and developers focused on advancing end-to-end driving systems, offering a streamlined experience for competition engagement and progress monitoring.

EgoGPT is an AI tool designed to process and analyze video content, providing a comprehensive textual description of the events within. Users can upload a video file, and the application will generate a detailed report outlining visual observations, actions performed, and audio elements, including a full transcription of any speech present. This tool is built on the EgoLife paper and is available as a Hugging Face Space, indicating its potential use for research and development in natural language processing and video understanding. While currently experiencing runtime errors, its core functionality aims to offer deep insights into video content through AI analysis.

Dreambooth is an AI tool available on Hugging Face, designed for customizing and training AI models. It enables users to personalize models by integrating specific subjects, making it valuable for various applications in AI image generation and manipulation. The tool is particularly useful for research, educational purposes, and personal projects where tailored AI models are required. While the current live website indicates a runtime error requiring a write token for login, its core functionality is centered around advanced model customization.

Florence 2 Large Ft is an AI tool developed by SixOpen, available as a Hugging Face Space, designed for comprehensive image and video analysis. It offers capabilities such as generating captions, detecting objects within visual media, and performing Optical Character Recognition (OCR). Users can upload an image or video, select a specific task, and optionally provide text input to enhance the analysis. This tool is suitable for various applications requiring detailed visual data interpretation. However, please note that the Space is currently paused, and users need to contact the author to restart it.

Gradio OpenAI CLIP Grad-CAM is a tool designed for visualizing the decision-making process of artificial intelligence models, specifically focusing on image-based predictions. It integrates Gradio for the user interface, OpenAI CLIP for understanding image-text relationships, and Grad-CAM for generating visual explanations of model predictions. This combination allows users to gain insights into which specific regions or features within an image are most influential in a model's output. The tool is particularly valuable for educational purposes, helping students and practitioners understand complex AI behaviors, and for researchers who need to analyze and debug model performance by observing its internal reasoning.

Gradio Demo Streaming is a Hugging Face Space designed for showcasing AI model capabilities through real-time data streaming and interactive applications. While currently paused, it offers a platform for developers and researchers to deploy and share their Gradio-based AI demos. The tool is ideal for educational purposes, allowing users to interact with AI models directly and observe their performance. Its integration within the Hugging Face ecosystem suggests a focus on community collaboration and easy deployment of machine learning applications. Users interested in utilizing this Space are encouraged to reach out to the author to request a restart, indicating its potential for future use in AI demonstrations and interactive projects.

Guanaco Playground Tgi is an AI chatbot tool hosted on Hugging Face Spaces by uwnlp, designed for interacting with the Guanaco 33B language model. This application allows users to input any question or comment and receive a comprehensive and courteous response. A key feature is the ability to fine-tune the response style using adjustable sliders, offering a degree of control over the AI's output. It serves as a playground for experimentation and research with advanced conversational AI.

Grok 1 Test is a demonstration space hosted on Hugging Face, specifically designed for users to interact with and test the Grok-1 model. Built using Gradio, it offers a direct interface for exploring the model's functionalities. While the current live website indicates a runtime error, the intention is to provide a platform where individuals can experiment with the Grok-1 model and understand its capabilities. This tool is ideal for those interested in AI model testing and development, offering a hands-on experience with a specific large language model.

Black in AI is a non-profit organization dedicated to increasing the presence and inclusion of Black individuals in the field of Artificial Intelligence. It aims to shift power dynamics within the AI ecosystem by fostering a barrier-free environment where community members can contribute and accelerate their work. The organization offers various programs including community engagement, education and career development, and initiatives focused on civil society, policy, research, and entrepreneurship. Black in AI supports its members through workshops, conferences, grants, and networking opportunities, connecting them with industry leaders and academic institutions. It serves as a vital resource for career advancement and community building within the AI space.

LiteratureDL4Graph offers a comprehensive and organized collection of recent academic papers focused on deep learning for graphs. This resource is invaluable for researchers and practitioners in the field, providing a curated list of literature sorted by topic and venue. Users can explore various sub-topics including Node Representation Learning (unsupervised, heterogeneous, dynamic graphs), Knowledge Graph Embedding, Graph Neural Networks, and diverse applications across NLP, Computer Vision, Recommender Systems, and Bioinformatics. The collection details key papers with their authors, publication venues, and relevant keywords, making it an essential tool for literature review and staying current with advancements in graph deep learning.