Projects

### Overview: ArtificialCast is a type-safe transformation utility for C# that leverages inference and AI to seamlessly convert between strongly typed objects. It simplifies the process of casting and transforming types without the need for manual mapping or boilerplate code. ### Key Features: - **Zero Config**: Just define your types. - **Bidirectional Casting**: Cast any type to any other. - **Schema-Aware Inference**: Auto-generates JSON Schema for the target type. - **LLM-Powered Transformation**: Uses AI to "fill in the blanks" between input and output. - **Testable & Deterministic-ish**: Works beautifully until it doesn't. ### Tech Stack: C# ### Why It Matters: ArtificialCast makes complex type conversions and transformations straightforward and efficient, eliminating the need for manual mapping and boilerplate code. It leverages AI to infer and transform types, making it a powerful tool for developers looking to streamline their development processes and handle complex data conversions.

### Overview: Türkçe Yapay Zeka Terimleri projesi, profesyonel geliştirme ve yapay zeka alanlarında kullanılan Türkçe terimleri içermektedir. ### Key features: - Türkçe yapay zeka terimleri ve terimler sözüğü temel alınarak oluşturulmuştur. - Değerli hocamıza teşekkür ederiz. - **TÜBA** ve **TBD** kendi sözlükleri var internet üzerinden faydalanabilir. - **SİU** kurultaylarını biliyorsunuz ve onlarda yakın zamanlı terimler için karşılıklar kullanabilirsiniz. - **Türkçe Yapay Zeka Terimleri** projesi güzel bir fikir ve kitabımın arkasındaki sözlüğü temel almanızda hiçbir sorun yok. - **Türkçe Yapay Zeka Terimleri** projesi için biraz yardımcı olmak için katkıda bulunmuştur. ### Tech stack: Python ### Why it matters: Bu proje, profesyonel geliştirme ve yapay zeka alanlarında kullanılan Türkçe terimleri içermektedir. Bu, geliştirme projelerinde kullanılan terimleri kolayca bulabilirsiniz. Ayrıca, bu proje, profesyonel geliştirme ve yapay zeka alanlarında kullanılan Türkçe terimleri içermektedir.

Overview: This repository contains code, data, prompts, and results related to the semi-automatic pipeline of Ontology and Knowledge Graph Construction using four different Large Language Models (LLMs): Mixtral 8x22B Instruct v0.1-Instruct-v0.1, GPT-4o, GPT-3.5, and Gemini. The project explores the potential of utilizing LLMs to generate Knowledge Graphs (KGs) and showcases the feasibility of a semi-automated pipeline by creating a KG on deep learning methodologies using scholarly publications. Key features: - Competency Questions (CQs) generation - Ontology development based on CQs - Knowledge Graph construction using the ontology - Evaluation of the resultant KGs with minimal human expert involvement Tech stack: Python Why it matters: This repository provides a practical demonstration of the potential of utilizing LLMs for the semi-automatic construction of Ontologies and Knowledge Graphs, showcasing the feasibility of a pipeline that leverages SOTA LLMs for knowledge extraction and graph construction. Repository: fusion-jena/automatic-KG-creation-with-LLM Primary language: Python GitHub stars: 365 GitHub description: Automatic Ontology and Knowledge Graph construction with LLM README excerpt:

### Overview: MathVista is a comprehensive project that evaluates mathematical reasoning in visual contexts. It includes datasets, models, and tools for assessing the performance of foundational models. ### Key Features: - **MathQA**: Evaluates mathematical reasoning through question answering tasks. - **Mathematical Reasoning**: Focuses on the evaluation of mathematical reasoning. - **Multi-Modal**: Incorporates multiple modalities for enhanced understanding. - **ScienceQA**: Provides datasets for evaluating mathematical reasoning in scientific contexts. ### Tech Stack: Jupyter Notebook ### Why it Matters: MathVista offers a platform for researchers and developers to evaluate the performance of mathematical reasoning models in various contexts, including visual and scientific scenarios. This project is crucial for advancing the field of mathematical reasoning and improving the accuracy and effectiveness of AI-driven solutions. Repository: lupantech/MathVista GitHub Stars: 367 GitHub Description: MathVista: data, code, and evaluation for Mathematical Reasoning in Visual Contexts README Excerpt: # MathVista: Evaluating Math Reasoning in Visual Contexts For more details, please refer to the project page with dataset exploration and visualization tools: [https://mathvista.github.io/](https://mathvista.github.io/). :bell: If you have any questions or suggestions, please don't hesitate to let us know. You can comment on the [Twitter](https://twitter.com/lupantech/status/1717313355780964608), or post an issue on this

Overview: ChemDataExtractor is a toolkit designed to automatically extract chemical information from scientific documents. Key features: - HTML, XML, and PDF document readers - Chemistry-aware natural language processing pipeline - Chemical named entity recognition - Rule-based parsing grammars for property and spectra extraction - Table parser for extracting tabulated data - Document processing to resolve data interdependencies Tech stack: Python Why it matters: ChemDataExtractor simplifies the process of extracting chemical information from scientific literature, making it easier for researchers and scientists to analyze and understand chemical data from various sources.

### Overview: StyleLLM 文风大模型是一个基于大语言模型的文本风格迁移项目，利用大语言模型学习指定文学作品的写作风格，生成带有特定风格的文本。 ### Key features: - 文本风格迁移 - 四大名著风格模型 - 量化模型列表（AWQ，4bits） ### Tech stack: - Python ### Why it matters: 通过StyleLLM，用户可以将指定风格的文本移植到其他通用文本中，实现文字修饰、润色或风格模仿的效果，适用于各种文本风格的转换和风格化聊天。

Overview: The `lmqg` library provides a Python framework for generating questions and answers (QAG) using language models (LMs). It supports nine languages and includes generation in one line of code, model training and evaluation, hosting QAG models, and an online web service for generating questions and answers. Key features: - Generation in one line of code for nine languages - Model training and evaluation - Hosting QAG models on a web application or REST API - AutoQG service for online question and answer generation Tech stack: Python Why it matters: The library simplifies the process of creating QAG models, making it accessible for developers and researchers. It supports a wide range of languages and provides robust tools for training, evaluating, and hosting models.

### Overview: Model Studio CLI, an official command-line interface for AI Agent frameworks, simplifies the process of exposing models, searching, multimodal capabilities, and workflow management. ### Key Features: - **Model Generation**: Full-modality generation across text, image, video, and speech, with editing and reference-based generation. - **Asset Understanding**: Parse and ask questions about images, documents, audio, and long videos. - **App Orchestration**: Call Managed Agents, agents, and workflows published on Aliyun Model Studio, integrated with knowledge bases, memory, web search, and MCP tools. - **Native MCP Setup**: Register Bailian MCP servers in Codex, Claude Code, Cursor, Qoder, Qoder Work, QwenWork (千问办公), Qwen Code, Gemini CLI, or Pi. - **Training & Deployment**: Validate and upload datasets, fine-tune models, deploy dedicated models as endpoints. - **Account Operations**: Login, UI-based configuration, model marketplace, usage and quota management, rate-limit increases, team seat management. - **Plan Onboarding**: Connect subscription plans such as Token Plan to the CLI and common coding agents in one step. ### Tech Stack: TypeScript ### Why It Matters: Simplifies the process of exposing models, searching, multimodal capabilities, and workflow management, making AI Agent frameworks more accessible and easier to use.

### Overview: This repository serves as a curated list of resources related to Semantic Search and Semantic Similarity tasks, focusing on advancements in the field of natural language processing and information retrieval. ### Key Features: - Papers: A collection of seminal papers from various years, covering a wide range of topics including semantic textual similarity, information retrieval, and more. - Articles: A selection of articles discussing the practical applications and challenges of semantic search. - Libraries and Tools: Resources for implementing semantic search algorithms and tools. - Datasets: Useful datasets for testing and evaluating semantic similarity models. - Milestones: A chronological overview of significant developments in the field. ### Tech Stack: The repository uses Markdown for documentation and includes links to various papers, articles, and datasets. ### Why it Matters: Semantic search and similarity tasks are crucial in modern information retrieval systems, enabling more accurate and context-aware search results. This repository provides a valuable resource for researchers, developers, and practitioners looking to stay updated with the latest advancements in this field.

### Overview: This repository is a curated list of projects related to Natural Language Processing (NLP) that make a geek smile for their existence. It features projects that are directly connected or useful for NLP, inspired by Joseph Misiti's GitHub project "awesome-machine-learning#python". ### Key Features: - **Projects**: A collection of projects related to NLP. - **Frameworks**: Resources and frameworks for NLP. - **Machine Learning**: Projects related to machine learning. - **Language Modelling**: Projects focusing on language modeling. - **Other ML**: Projects related to other machine learning techniques. - **Other Cool stuff**: Additional projects that are not specifically related to NLP but are still interesting. - **Contribute**: Instructions on how to contribute to the repository. - **License**: The repository is licensed under the MIT License. ### Tech Stack: Python ### Why it Matters: These projects showcase the diversity and creativity in NLP, making them a joy to explore and contribute to. They are curated to provide a comprehensive overview of NLP resources and frameworks, making them valuable for developers and researchers alike.

### Overview: CM3Leon is an open-source implementation of an all-new multi-modal AI model that uses an autoregressive decoder to generate both text and images. The model is designed to revolutionize automation by leveraging a large, diverse multimodal dataset and augmented retrieval pretraining. It includes a contrastive decoding technique to enhance the quality of generated samples. ### Key Features: - **Text and Image Generation**: CM3Leon is capable of generating both text and images using an autoregressive decoder. - **Pretraining and Instruction Tuning**: The model is trained in two stages, first using a large dataset and then fine-tuned with instruction tuning. - **Contrastive Decoding**: Enhances the quality of generated samples through contrastive decoding. ### Tech Stack: - **Programming Language**: Python ### Why It Matters: CM3Leon represents a significant advancement in multi-modal AI, promising to transform industries by integrating text and image generation capabilities. This model could revolutionize automation and various sectors by providing smarter and more efficient solutions. Repository: kyegomez/CM3Leon GitHub Stars: 365 GitHub Description: An open source implementation of "Scaling Autoregressive Multi-Modal Models: Pretraining and Instruction Tuning," an all-new multi-modal AI that uses just a decoder to generate both text and images.

### Overview: HackBot is an AI-powered cybersecurity chatbot designed for providing helpful and accurate answers to cybersecurity-related queries. It also performs code analysis and scan analysis, making it a versatile tool for cybersecurity students and professionals. ### Key Features: - **Multi-Agent Orchestration**: Plans and orchestrates specialist agents in parallel, including researchers, a SARIF analyst, a document analyst, and a tutor. - **Model Context Protocol (MCP)**: Each tool operates in a separate MCP server process, and agents are MCP clients. - **Agent Communication Protocol (ACP)**: Agents are published by `hackbot-acp` and can be discovered and called from any ACP client. - **Any File or Folder as Input**: Reviews SARIF reports and any other text file or folder, including a Prowler/Nmap output directory. - **Interactive Terminal App**: Built with Rich, featuring a dashboard, chat with streaming answers, live activity tree, result and sources windows, and a file viewer with search and triage review capabilities. - **Research and Defense Only**: The system prompts for refuse exploit/malware generation, and it uses the cheapest OpenAI model by default (`gpt-5.6-luna`). ### Tech Stack: - **Python**: The primary language used for developing HackBot. ### Why It Matters: HackBot is a practical demonstration of an agentic AI system, showcasing how multiple agents can work in parallel to solve complex cybersecurity problems. It provides a comprehensive solution for cybersecurity students and professionals, offering a step-by-step approach to cybersecurity research and defense.

### Overview: This repository contains a comprehensive summary of the most significant NLP highlights from 2018, organized into various categories such as AI ethics, research publications, trends, education, and more. The report aims to provide a guide for researchers and developers interested in the field, offering insights into the year's advancements and trends. ### Key Features: - **NLP 2019 Highlights**: A summary of the most important NLP stories from 2019, focusing on machine learning and NLP. - **NLP 2018 Highlights**: A collection of the year's most notable NLP stories, including research publications, trends, education, and more. - **Categories**: Categorized into AI ethics, research publications, trends, education, resources, industry, and other relevant topics. - **GitHub Stars**: 365 - **GitHub Description**: The most important NLP highlights of 2018 (PDF Report) ### Tech Stack: N/A ### Why It Matters: This report serves as a valuable resource for researchers and developers interested in the field of natural language processing, providing insights into the year's advancements and trends. The categorized format makes it accessible to a diverse audience, including those new to the field and experts looking to stay updated.

Overview: ARIA is a Windows desktop application designed to turn system audio into live subtitles, allowing users to listen to audio from various sources and show recognized text in a movable overlay. It supports real-time and offline translation for enhanced functionality. Key features: - Real-time subtitles - Offline translation - Windows Live Captions mode - NVIDIA GPU support for precise mode - Local speech recognition for precise mode Tech stack: Python Why it matters: ARIA simplifies the process of real-time subtitles for Windows users, making it easy to listen to audio from video players, browsers, games, and calls, and show recognized text in a movable overlay. The app supports offline translation and is optimized for Windows 11 Live Captions, making it a valuable tool for real-time captioning in various audio sources. Repository: sayksii/Aria Primary language: Python GitHub stars: 362 GitHub description: ARIA - AI Realtime Intelligent Audio | Universal real-time AI subtitles for Windows

Overview: This tutorial provides a comprehensive guide to implementing a sequence labeling model using PyTorch, focusing on a task-aware neural language model. The model is designed to enhance sequence labeling tasks by training them concurrently with language models, making it particularly effective and adaptable. Key features: - Implementing a task-aware neural language model - Training sequence labeling concurrently with language models - Using PyTorch for implementation - Assumed knowledge of PyTorch and recurrent neural networks Tech stack: PyTorch Why it matters: This tutorial offers a practical approach to sequence labeling, leveraging advanced concepts like task-aware language models and conditional random fields (CRFs) to improve model performance and adaptability across various domains.

Overview: This repository contains a curated list of papers focused on the development and evaluation of language model agents using executable language grounding techniques. The primary goal is to facilitate interaction between language models and real-world environments, such as databases, web applications, and the physical world. Key features: - **Executable Language Grounding (XLANG)**: Focuses on building language model agents that can transform language instructions into executable code or actions. - **Real-World Applications**: Includes papers on integrating language models with databases, web applications, and robotic systems. - **Techniques**: Incorporates advancements like LLM + external tools, code generation, semantic parsing, and dialog systems. Tech stack: Python, JavaScript, Markdown, and GitHub. Why it matters: This repository serves as a valuable resource for researchers and developers interested in advancing the field of language model agents, enabling them to interact more effectively with real-world environments through conversational language models.

### Overview: A-RAG: Agentic Retrieval-Augmented Generation via Hierarchical Retrieval Interfaces Frontier language models have shown impressive reasoning and long-horizon tool-use capabilities. However, existing RAG systems still rely on two paradigms: Graph RAG and Workflow RAG. Neither paradigm allows the model to participate in retrieval decisions, limiting scalability with model improvements. ### Key Features: - **Hierarchical Retrieval Interfaces**: A-RAG exposes these interfaces directly to the model, enabling dynamic retrieval strategies. - **Keyword Search**: Facilitates quick retrieval of specific keywords. - **Semantic Search**: Enables semantic-based retrieval for more context-aware results. - **Chunk Read**: Supports multi-hop reasoning by reading and integrating chunks of information. ### Tech Stack: Python A-RAG is built using Python, leveraging the `uv` framework for easy integration and management. ### Why It Matters: A-RAG is a truly agentic framework that leverages the strengths of frontier language models, enabling dynamic retrieval strategies, iterative execution, and interleaved tool use. This makes it a game-changer for scaling reasoning capabilities in language models. ### Repository: Ayanami0730/arag - **GitHub Stars**: 363 - **Primary Language**: Python - **GitHub Description**: A-RAG: Agentic Retrieval-Augmented Generation via Hierarchical Retrieval Interfaces. State-of-the-art RAG framework with keyword, semantic, and chunk read tools for multi-hop QA. - **GitHub Description**: If you find our project helpful, please give us a

Overview: DiffPoseTalk is a project that leverages diffusion models to generate diverse and stylistic 3D facial animations and head pose generation based on speech-driven input. The system aims to create realistic and expressive lip-synced animations. Key features: - Pretrained models for inference - Training and inference code - Raw, processed, and TF-HP dataset release Tech stack: Python Why it matters: This project showcases the application of diffusion models in generating realistic animations, making it valuable for both research and practical use in the field of computer graphics and AI. Repository: DiffPoseTalk/DiffPoseTalk Primary language: Python GitHub stars: 364 GitHub description: DiffPoseTalk: Speech-Driven Stylistic 3D Facial Animation and Head Pose Generation via Diffusion Models

### Overview: Car Simulator is an autonomous car simulation project utilizing JavaScript and WebGL. It employs fuzzy control systems, genetic algorithms, and particle swarm optimization to simulate car movements and optimize parameters for better performance. ### Key Features: - **Autonomous Car Simulation**: Simulates the movement of an autonomous car on a map. - **VR Support**: Compatible with VR devices for immersive experience. - **Fuzzy Control System**: Uses custom fuzzy rules and discrete center of gravity defuzzifier. - **Genetic Algorithm**: Trains radial base function networks (RBFN) using real-valued genetic algorithms. - **Particle Swarm Optimization**: Trains RBFN using particle swarm optimization (PSO). - **4-Dimensional Output Data Format**: Outputs the rotation angle of steering wheel based on 4D input data. - **6-Dimensional Output Data Format**: Outputs the rotation angle of steering wheel based on 6D input data. - **VR Preview**: Provides a preview of the simulation with VR support. - **Operating Instructions**: Detailed instructions for starting the simulation, setting parameters, and saving trajectory data. ### Tech Stack: - JavaScript - WebGL ### Why It Matters: Car Simulator offers a comprehensive simulation environment for autonomous car enthusiasts and developers, allowing them to experiment with different control systems and optimization techniques. The project's focus on real-world applications and advanced algorithms makes it valuable for both educational and practical purposes. ### Repository: - Repository: Jasonnor/Car-Simulator - GitHub Stars: 364 - GitHub Description: Autonomous car simulator (based on JavaScript & WebGL)

### Overview: DocDiff 是一个用于文档增强的轻量级扩散模型，能够处理文档去模糊、去噪、文档二值化、去水印和印章等任务。它在128x128分辨率上以Batchsize=64训练只需要12GB显存。此外，DocDiff还包含1597个红色系印章及其对应的二值化掩膜，适用于印章合成和消除等任务。 ### Key features: - **文档增强模型**：DocDiff是一个轻量级的基于残差预测的扩散模型，适用于文档去模糊、去噪、文档二值化、去水印和印章等任务。 - **高分辨率处理**：在128x128分辨率上以Batchsize=64训练，仅需12GB显存。 - **丰富的数据集**：包含1597个中文场景下的红色系印章及其对应的二值化掩膜，适用于印章合成和消除等任务。 - **多种应用场景**：不仅限于文档增强，还可以应用于其他img2img任务，如自然场景去模糊、去噪、超分、图像修复等低级任务，以及语义分割等高级任务。 ### Tech stack: Python ### Why it matters: DocDiff是ACM Multimedia 2023的论文[DocDiff: Document Enhancement via Residual Diffusion Models](https://dl.acm.org/doi/abs/10.1145/3581783.