Research & Education

Eye On A.I. is a dedicated platform offering a unique blend of news, insightful analysis, and critical data within the rapidly evolving artificial intelligence sector. It serves as a valuable resource for staying informed on the latest developments and trends in AI. The platform features a podcast that includes discussions with leading AI authorities, such as Professor Mausam from IIT Delhi, providing in-depth perspectives on the global AI landscape, including comparisons between India, the US, and China. Transcripts of these discussions are also available for download, allowing users to delve deeper into the expert insights. Eye On A.I. aims to provide a comprehensive understanding of the challenges and opportunities within the AI domain.

T3Bench is the first comprehensive benchmark specifically designed for evaluating current progress in text-to-3D generation models. It includes a diverse set of 300 text prompts categorized into three increasing complexity levels. To provide a thorough assessment, T3Bench proposes two automatic metrics: a quality metric and an alignment metric. The quality metric combines multi-view text-image scores and regional convolution to detect quality and view inconsistency in generated 3D content. The alignment metric utilizes multi-view captioning and Large Language Model (LLM) evaluation to measure the consistency between the input text and the 3D output. Both metrics have been shown to closely correlate with different dimensions of human judgments, offering an efficient paradigm for evaluating text-to-3D models. The benchmark also provides mesh results for various prompt sets and methods, making it a valuable resource for researchers and developers in the field.

simple_GRPO is an open-source implementation of the GRPO algorithm, specifically designed for reproducing r1-like LLM thinking. It utilizes a core loss calculation formula referenced from Hugging Face's trl, but with a significantly simplified codebase. The tool aims to save GPU memory, enabling feasible and efficient training, and helps users quickly understand and experiment with Reinforcement Learning processes like GRPO. It supports features such as improved multi-answer generation, regrouping, penalty on KL, and parameter tuning, all within approximately 200 lines of code across two files. The reference model is decoupled, allowing it to run on separate GPUs, which prevents multiple copies from being created by torch’s multiprocessing and enables training of large models on less powerful hardware.

text_mining_resources is a comprehensive, curated list of resources designed for individuals interested in learning about natural language processing (NLP), text analytics, and working with unstructured data. The repository offers a wide array of materials, including books, blogs, and articles, covering fundamental and advanced topics. Users can find resources on biases in NLP, data scraping, text cleaning, stemming, dimensionality reduction, sarcasm detection, document classification, entity extraction, topic modeling, sentiment analysis, and more. It also includes sections on major NLP conferences, online courses, APIs, libraries, and datasets, making it a valuable hub for students, researchers, and practitioners in the field.

tiny-diffusion offers a character-level language diffusion model for text generation, implemented in just 365 lines of Python code. This compact model, with 10.7 million parameters, is trained on Tiny Shakespeare, making it suitable for local experimentation and learning. The repository also features a tiny GPT implementation in 313 lines, with significant code overlap between the two models. It supports parallel decoding for diffusion and autoregressive generation for GPT. Users can train both models from scratch, visualize the generation process, and compare the diffusion and GPT models side-by-side. The diffusion model introduces key modifications like a mask token, bidirectional attention, confidence-based parallel decoding, and a training objective focused on unmasking.

tomesd is an open-source Python and PyTorch-based tool designed to accelerate Stable Diffusion models by implementing Token Merging (ToMe). This technique reduces computational load by merging redundant tokens within the transformer blocks, leading to faster image generation and lower memory consumption. tomesd works out-of-the-box with various Stable Diffusion models, including v1, v2, Latent Diffusion, and Diffusers, and does not require additional training. While it's a lossy process, it minimizes quality degradation while providing substantial speed and memory benefits. It can be applied to existing Stable Diffusion environments and is compatible with other efficient transformer implementations like xformers.

WebArena is a self-hostable, open-source web environment designed for building and evaluating autonomous AI agents. It provides a realistic web environment, enabling researchers and developers to reproduce results from academic papers and conduct new experiments. The platform has been significantly enhanced by AgentLab, offering features like parallel experiments using BrowserGym, integration of popular web navigation benchmarks such as VisualWebArena, and a unified leaderboard for reporting results. It also includes improved handling of environment edge cases, making it a robust framework for developing and testing AI agents in complex web interactions. The repository provides detailed instructions for installation, environment setup, and end-to-end evaluation, including generating test data and launching evaluations with various reasoning agents.

Trankit is a light-weight, transformer-based Python toolkit designed for multilingual Natural Language Processing (NLP). It offers a trainable pipeline for fundamental NLP tasks across more than 100 languages, and includes 90 downloadable pretrained pipelines for 56 languages. Trankit outperforms other state-of-the-art multilingual toolkits like Stanza in various tasks, including sentence segmentation and dependency parsing, while maintaining efficiency in memory usage and speed. Key features include an Auto Mode for automatic language detection, a command-line interface for ease of use, and support for tasks such as tokenization, part-of-speech tagging, morphological feature tagging, dependency parsing, and named entity recognition. It also allows users to build and share customized pipelines.

The uvadlc_notebooks repository offers a comprehensive collection of Jupyter notebook tutorials specifically designed for the Deep Learning Course at the University of Amsterdam (MSc AI). These notebooks aim to bridge the gap between theoretical concepts and practical implementation, covering diverse topics such as optimization techniques, transformers, graph neural networks, and more. The materials are available for both Fall 2023 and Fall 2024 course editions, with support for PyTorch and PyTorch Lightning, as well as JAX+Flax. Users can run the notebooks locally on CPU, utilize Google Colab for GPU access, or leverage the Snellius cluster for larger-scale training. The tutorials are integrated into PyTorch Lightning's official documentation, making them a valuable resource for students and practitioners alike.

BobaAI functions as an AI co-pilot designed for generative ideation, assisting users with various research and strategic planning tasks. It can search the web for articles and news to answer qualitative research questions, providing summaries and deeper analysis of findings. The tool also facilitates brainstorming by exploring opportunities, threats, and probabilities of future scenarios. Furthermore, BobaAI helps in developing strategies using frameworks like 'Playing to Win' and can generate 'where to play' and 'how to win' choices. It also supports storyboarding, allowing users to create illustrated scenes and customize scripts for current or future state scenarios, making it a comprehensive tool for ideation and strategic development.

KLETech-Center of Excellence in Visual Intelligence (CEVI) is a dedicated research center focused on the advancement of computer vision and immersive technologies. The center places a strong emphasis on 3D vision and aims to create solutions with significant real-world impact. CEVI actively conducts workshops, fosters collaborations with industry partners, and engages with other research institutes to drive innovation. Its work is centered around artificial intelligence and data engineering, primarily through funded projects, with a core mission to strengthen partnerships between academia and industry. This focus allows CEVI to contribute to both theoretical advancements and practical applications in visual intelligence.

OpenManus-RL is an open-source initiative, collaboratively led by Ulab-UIUC and MetaGPT, dedicated to advancing reinforcement learning (RL) tuning for large language model (LLM) agents. Inspired by successful RL tuning in models like Deepseek-R1, this project explores novel algorithmic structures, diverse reasoning paradigms, and sophisticated reward strategies. It supports rigorous testing on agent benchmarks such as GAIA, AgentBench, WebShop, and OSWorld, with all progress and tuned models openly shared. The platform integrates advanced RL algorithms like PPO and DPO through the Verl submodule, offering efficient and flexible training capabilities. It also provides a simplified library for Supervised Fine-Tuning (SFT) and GRPO tuning, making it a comprehensive solution for researchers and developers looking to push the boundaries of agent reasoning and tool integration.

Echo Reading is an AI-powered eBook reader designed to enhance the reading experience by integrating annotation and AI chat functionalities directly into the platform. It allows users to interact with their PDF documents in a novel way, enabling them to select text and instantly ask questions without the need for manual copy-pasting. This open-source tool prioritizes user privacy and data security by utilizing the user's own OpenAI API key, ensuring that all data remains local. It aims to streamline the process of understanding complex texts and conducting research by providing immediate AI assistance within the reading environment.

RE•WORK is a leading global events company specializing in AI and Deep Learning. They create and organize globally renowned summits, workshops, and dinners, bringing together the brightest minds in AI from both industry and academia. Their events showcase the latest technological advancements in AI, alongside practical examples for applying AI to solve challenges in business and society. RE•WORK fosters a diverse AI community, featuring speakers from tech giants like Google, Apple, DeepMind, Amazon, and Facebook, as well as world-renowned academic institutions. Beyond events, RE•WORK provides analysis of current trends through podcasts, white papers, and video interviews, and hosts initiatives like the Women in AI Dinner and Rising Stars Sessions.

This repository, A-Curated-List-of-ML-System-Design-Case-Studies, offers a comprehensive collection of over 300 machine learning (ML) system design case studies from more than 80 leading companies. It details practical applications and insights into how ML is used to improve products and processes across various industries like tech, finance, healthcare, and e-commerce. The case studies cover diverse ML applications such as computer vision, natural language processing, recommender systems, and fraud detection. Each study is sourced from detailed blogs, papers, or articles, providing authentic and in-depth information on model designs, evaluation criteria, and deployment architectures. It's a valuable resource for anyone looking to understand real-world ML systems in production.

AnglE is an open-source library designed for training and inferring state-of-the-art BERT/LLM-based sentence embeddings. It utilizes an angle-optimized approach, offering various loss functions like AnglE loss, Contrastive loss, CoSENT loss, and Espresso loss. The library supports both BERT-based and LLM-based models, including bi-directional LLMs, and facilitates single-GPU and multi-GPU training. AnglE has achieved SOTA performance on benchmarks like STS and MTEB, with models trained using AnglE reaching top positions. It provides a flexible framework for researchers and developers to build and deploy high-quality sentence embedding models.

This open-source project offers a comprehensive collection of PyTorch implementations for various deep learning papers and algorithms. Each implementation is meticulously documented with side-by-side explanations, designed to enhance understanding of complex concepts. The repository is actively maintained, with new implementations added almost weekly, covering a wide range of topics including transformers (original, XL, Switch, Feedback, ViT), optimizers (Adam, AdaBelief, Sophia), GANs (CycleGAN, StyleGAN2), and reinforcement learning (PPO, DQN). It also features implementations for Capsule Networks, Distillation, and various normalization layers. This resource is ideal for students, researchers, and developers looking to delve deeper into the practical application and theoretical underpinnings of deep learning.

The University of Waterloo's Data & Artificial Intelligence Institute (Waterloo.AI) is a multidisciplinary research institute dedicated to advancing AI for economic prosperity and quality of life. It focuses on developing intelligent systems for various applications, including disease detection, language understanding, and vehicle navigation. The institute actively collaborates with industry partners to bridge the gap between academic research and practical, deployable AI solutions. Waterloo.AI aims to foster innovation and talent in the AI field, contributing to real-world impact through its research and partnerships.

AlpacaEval is an automatic evaluator designed for instruction-following language models, providing a fast, cheap, and highly correlated alternative to human evaluation. It boasts a Spearman correlation of 0.98 with ChatBot Arena, costing less than $10 of OpenAI credits and running in under 3 minutes. The tool offers precomputed leaderboards for common models, an automatic evaluator validated against 20K human annotations, and a toolkit for building advanced automatic evaluators with features like caching, batching, and multi-annotators. It also includes 20K human evaluation data and a simplified AlpacaFarm evaluation dataset. AlpacaEval is particularly useful for rapid model development and iterative testing, though it cautions against replacing human evaluation for high-stakes decision-making due to potential biases and limitations in instruction representativeness.

Auto-Deep-Research is an open-source, fully-automated personal AI assistant designed as a cost-effective alternative to OpenAI's Deep Research. Built on the AutoAgent framework, it boasts high performance on the GAIA Benchmark and offers universal LLM support, seamlessly integrating with a wide range of models including OpenAI, Anthropic, Deepseek, vLLM, Grok, and Huggingface. The tool supports both function-calling and non-function-calling interaction LLMs and handles file uploads for enhanced data interaction. Users can get started instantly with a simple command, requiring zero configuration for an out-of-the-box experience. It aims to provide a personal assistant at a much lower cost, leveraging pay-as-you-go LLM API keys.

albert_zh is an open-source implementation of A Lite Bert for Self-Supervised Learning of Language Representations, specifically optimized for Chinese language processing. Based on the BERT architecture, ALBERT introduces improvements like factorized embedding parameterization and cross-layer parameter sharing, significantly reducing the number of parameters while retaining or even improving accuracy. This leads to faster training and inference times, making it suitable for real-time applications and resource-constrained environments. The repository provides various pre-trained ALBERT models for Chinese, including tiny, small, base, large, and xlarge versions, with options for TensorFlow, PyTorch, and Keras. It includes scripts for pre-training on custom data and fine-tuning on downstream tasks like semantic similarity prediction, with examples provided for the LCQMC dataset.

ASearcher is an open-source framework designed for large-scale online reinforcement learning (RL) training of search agents, aiming to advance Search Intelligence to expert-level performance. It provides model weights, detailed training methodologies, and data synthesis pipelines, making it fully committed to open-source development. Key features include a prompt-based LLM agent for autonomous QA pair generation, a fully asynchronous agentic RL framework that decouples trajectory collection from model training, and the ability to enable long-horizon search with tool calls exceeding 100 rounds. ASearcher achieves cutting-edge performance on challenging QA benchmarks like GAIA, xBench-DeepSearch, and Frames, demonstrating substantial improvements through RL training. It also offers comprehensive guidance for building and training customized agents.

attention_with_linear_biases is a GitHub repository offering the implementation of the Attention with Linear Biases (ALiBi) method for transformer language models. This method, presented in the ICLR 2022 paper 'Train Short, Test Long,' allows models to be trained on shorter input sequences (e.g., 1024 tokens) and then perform inference on significantly longer sequences (e.g., 2048 tokens or more) without requiring fine-tuning. The repository provides code and models for conducting experiments, specifically on the WikiText-103 dataset. ALiBi simplifies the positional encoding process by adding a linear bias to each attention score instead of using traditional position embeddings, which can improve performance even in non-extrapolating scenarios. The implementation details, including removing position embeddings and setting up the relative bias matrix, are clearly outlined.

Awesome-Image-Colorization is a comprehensive, open-source collection of deep learning-based research papers focused on image and video colorization. This GitHub repository serves as a valuable resource for researchers and developers interested in the field, offering direct links to academic papers, their corresponding source code, and demo programs. The collection covers a wide array of colorization methods, including automatic colorization, user-guided colorization (based on scribbles, reference images, palettes, or text), and video colorization. It is continuously updated with new research, making it an essential reference for staying current with advancements in AI-powered colorization.