#Datasets

Argilla is a collaborative platform for AI engineers and domain experts to create high-quality datasets. It focuses on data quality to enhance model training and evaluation, allowing efficient management and iteration on AI datasets and models. Argilla is applicable in AI fields such as NLP, LLMs, and multimodal systems. This community-driven project supports open-source datasets and models, enabling organizations like the Red Cross and Prolific to enhance their AI projects.

The repository offers a variety of annotated datasets for entity recognition and NER, covering domains such as news, medical, and finance. While updates stopped in 2020, it remains a valuable source for English-language datasets and supports format conversion to CoNLL 2003. Additionally, it connects to global datasets, providing a resource for multilingual NER study. Contributions through issues or pull requests are accepted to enrich this repository.

The project investigates various deep learning techniques to enhance emotional expression in Text-to-Speech systems. Focusing on Tacotron and DCTTS models, it explores fine-tuning strategies using datasets such as RAVDESS and EMOV-DB to augment speech naturalness and emotional depth. The research involves optimizing model parameters, applying novel training methodologies, and utilizing transfer learning in low-resource settings. The repository offers insights into utilizing neural networks for generating emotionally nuanced speech, along with practical implementations and evaluations of diverse methods.

Investigate a benchmark for parameter-efficient transfer learning in computer vision, assessing 25 leading algorithms on 30 varied datasets. The platform provides a modular codebase for comprehensive analysis in image recognition, video action recognition, and dense prediction. Pre-trained models like ViT and Swin are used to attain high performance with fewer parameters. The benchmark facilitates easy evaluation and continuous updates for new PETL methods and applications.

The repository offers a detailed collection of datasets, codes, and scholarly articles related to Multi-Task Learning from a machine learning perspective. It covers extensive topics such as benchmarks in domains like computer vision, NLP, robotics, and recommendation systems. The project also delves into various architectures including hard and soft parameter sharing, and explores optimization techniques essential for advancing MTL. Open to contributions, this platform aims to support collaboration and improve multi-tasking methodologies, facilitating researchers and industry experts in addressing complex challenges effectively.

Access an extensive collection of open-source datasets for instruction tuning, suitable for training both text and multi-modal chat-based large language models (LLMs) like GPT-4, ChatGPT, LLaMA, and Alpaca. This repository includes visual-instruction, text-instruction, and RLHF datasets, offering crucial resources for LLM fine-tuning and development. It provides multilingual and multi-task datasets created from both human and machine sources, which facilitate specific task solutions. Leverage these datasets and a comprehensive codebase to advance LLM research and development.

This collection offers diverse tools, datasets, and resources for Bangla computing, catering to those working on Natural Language Processing (NLP) in Bangla (Bengali). It features various typing tools, phonetic parsing libraries, language-processing datasets, and NLP resources such as POS taggers and sentiment analysis. Additionally, it includes machine translation, OCR, and TTS resources, advancing capabilities in the field. These resources encourage innovation and community participation within the Bangla language technology sphere.

CLUE provides a robust framework for evaluating Chinese language understanding, including diverse datasets, baseline models, and leaderboards. It supports a range of tasks with different complexities and scales, offering a testing ground for Chinese language models. The latest update introduces valuable resources such as the SuperCLUE benchmark and compatibility with PaddleNLP. As a foundational tool, CLUE facilitates the growth of Chinese language models through targeted evaluations. It supplies crucial tools and instructions for performing classification and reading comprehension tasks, and its corpus of over 14GB aids in pre-training and language modeling tasks, contributing to NLP advancements in academic and industrial fields.

This project introduces diffusion autoencoders that focus on meaningful and decodable image representation. Featured in CVPR 2022, it offers practical tools like Colab walkthroughs and web demos for sample generation, manipulation, and interpolation. The comprehensive documentation and LMDB datasets support ease of use. It also provides training and evaluation scripts for datasets such as FFHQ and CelebAHQ, facilitating advancements in AI image processing, and supplying essential tools for researchers and developers.

SPIN uses a self-play mechanism to improve language models, enabling self-enhancement through iteration. It generates training data from past iterations to refine model strategies and excels over models trained via direct preference optimization. SPIN achieves enhancements without needing extra human-annotated data beyond what's required for supervised fine-tuning. The method is theoretically grounded and validated on multiple benchmarks, ensuring data distribution alignment. Detailed setup guides and open-source availability aid replication and further exploration.

Discover a framework for constructing conversational Large Language Models tailored for healthcare applications. The project emphasizes the development of comprehensive instructional datasets utilizing NHS data, equipping researchers with tools to generate and train models with domain expertise. Access essential NHS UK Q/A and conversation datasets for research advancements. Installation is straightforward using pip, with additional specifications for LLaMA models. Access step-by-step tutorials for building a unique conversational model in healthcare. Delve into our discourse platform for ongoing updates on crafting specialized LLMs in the healthcare industry.

Explore a comprehensive selection of publicly available human preference datasets suitable for fine-tuning language models, reinforcement learning from human feedback, and assessment. Highlighted collections encompass OpenAI WebGPT Comparisons, OpenAI Summarization, and Anthropic Helpfulness and Harmlessness Dataset, among others. Offering resources aimed at NLP development, these datasets are derived from sources including Reddit, StackExchange, and ShareGPT, enriching understanding of human preferences in AI. They support the development of reward models and offer insights into evaluating human-generated content across varied fields, ideal for researchers and developers working on the advancement of language model alignment.

This repository provides a wide array of toolkits, datasets, and research papers focused on Simultaneous Translation, updated regularly to meet academic and professional demands. Notable tools such as Fairseq and SimulEval support sophisticated translation and evaluation processes. The repository includes datasets for various translation tasks, like text-to-text, speech-to-text, and speech-to-speech, featuring collections such as IWSLT, WMT, and MuST-C. It also offers a detailed paper list arranged by year and topic to facilitate understanding of field advancements. This repository is an essential resource for researchers and professionals working on Simultaneous Translation projects.

Explore comprehensive knowledge graph reasoning resources, including key papers, code, and datasets. Gain insights into neural, translational, and transformer models across static, temporal, and multi-modal graphs. Access transductive and inductive datasets, essential libraries, and stay updated on the latest advancements, supporting both academic and practical applications.

This resource offers a comprehensive collection of research papers, code repositories, tutorials, and datasets focused on lane detection technologies including 3D techniques and transformer methods. Featuring works from leading conferences like CVPR, ICCV, and ECCV, it is essential for those interested in advancing autonomous driving and road safety technologies.

An objective overview of instruction tuning methodologies, dataset structures, and model training in large language models. This survey evaluates applications across various domains and considers factors impacting outcomes, offering valuable insights into the challenges and future research directions.

LibFewShot provides broad support for various few-shot learning techniques, including non-episodic, meta-learning, and metric-based methods. It offers users a streamlined installation process, detailed documentation, and verified configuration for reliable results. Access renowned datasets such as Stanford Dogs and miniImageNet for research purposes. Designed for academic researchers, LibFewShot fosters community contributions and operates under the MIT License.

Explore a curated collection of resources dedicated to 3D hand pose estimation, featuring evaluation methodologies, groundbreaking research from arXiv, journals such as TPAMI and IJCV, and major conferences like CVPR and ICCV. Access robust datasets for depth and RGB-based analyses, and stay updated with the latest workshops and challenges in the field. This compilation serves as a key reference for researchers and developers focusing on innovations in 3D hand pose estimation through semi-supervised learning, advanced modeling, and real-time applications. Contributions to the repository are welcome to further enrich the collective knowledge base.

This compilation gathers advanced research materials on interaction-aware trajectory prediction, featuring datasets, academic papers, and public code useful across academia and industry. Covering insights into vehicle, pedestrian, and sports player scenarios, it is regularly updated by experts from Stanford University and UC Berkeley, facilitating collaboration. It also includes surveys on cutting-edge deep learning, neural networks, and autonomous driving technologies, with additional resources in reinforcement learning and decision-making, driving innovation forward.