Research & Education

X-VLA is the official implementation of "Soft-Prompted Transformer as Scalable Cross-Embodiment Vision-Language-Action Model," accepted to ICLR 2026. This open-source project introduces a soft prompt mechanism using embodiment-specific learnable embeddings to guide a unified Transformer backbone. This approach facilitates effective multi-domain policy learning across heterogeneous large-scale robot datasets. The resulting X-VLA-0.9B architecture demonstrates state-of-the-art generalization across six simulation platforms and three real-world robots, outperforming previous VLA methods in dexterity, adaptability, and efficiency. It supports a Server–Client architecture for distributed inference and offers various pre-trained models fine-tuned for specific robotic embodiments and benchmarks like AgiBot World Challenge, CALVIN, Google Robot, and LIBERO.

WindowsAgentArena (WAA) is a scalable Windows AI agent platform designed for testing and benchmarking multi-modal, desktop AI agents. It provides researchers and developers with a reproducible and realistic Windows OS environment, enabling the testing of agentic AI workflows across a diverse range of tasks. WAA supports the deployment of agents at scale using Azure ML cloud infrastructure, allowing for parallel execution of multiple agents and delivering quick benchmark results for hundreds of tasks in minutes. The platform includes features like a new difficulty mode for tasks, the Navi agent with Omniparser, and the open-sourced Omniparser screen understanding model. Users can deploy locally using Docker and WSL 2, or leverage Azure for parallel benchmarking.

WizardLM is a suite of large language models (LLMs) built upon the Evol-Instruct method, designed to enhance their ability to follow complex instructions. This project includes several specialized models: WizardLM focuses on general instruction following, WizardCoder excels in code generation, and WizardMath is optimized for mathematical reasoning. The models demonstrate strong performance against established benchmarks and even surpass some closed-source alternatives like ChatGPT 3.5 and Gemini Pro in specific tasks. WizardLM provides various model sizes, from 7B to 70B parameters, with different licensing options. It is a valuable resource for researchers and developers looking to leverage advanced open-source LLMs for a range of applications.

Elix, Inc. is an AI drug discovery company dedicated to "Rethinking Drug Discovery" by applying artificial intelligence to accelerate the process and maximize successful outcomes. The platform provides advanced predictive and generative models specifically designed for medicinal chemistry, offering an intuitive user interface. These AI models are trained on proprietary data from 16 pharmaceutical partners, covering critical aspects such as activity, ADME (absorption, distribution, metabolism, and excretion), and toxicity. Elix specializes in designing novel, non-obvious structures that have been proven effective across multiple drug discovery programs. The company also offers expert consulting and support from experienced PhD scientists, ensuring hands-on onboarding, seamless workflow integration, and ongoing assistance for its clients.

Reesi is an AI-powered platform designed to significantly improve patient access to oncology clinical trials by streamlining the search and recruitment process. It aggregates and structures publicly available trial data, integrating sponsor insights to provide a comprehensive and transparent overview. The platform helps oncologists, clinics, and study centers quickly identify relevant clinical trials, accelerating patient recruitment and access to potentially life-changing treatments. Reesi aims to foster a more connected and efficient ecosystem for clinical research in oncology, ensuring that suitable studies are found in seconds.

RegulatingAI is a non-profit organization dedicated to fostering responsible innovation in the field of artificial intelligence by focusing on AI regulations and legislation. It serves as a crucial bridge connecting AI experts, researchers, corporations, NGOs, academicians, students, and enthusiasts. The platform provides comprehensive resources including AI news, press releases, and detailed information on AI legislation and bill status. RegulatingAI actively promotes conversations between experts and policymakers to contribute to the development of ethical and responsible artificial intelligence laws. Users can access podcasts, event highlights, and stay informed about governmental hearings related to AI regulation in the U.S. Senate and Congress. The organization is committed to building a shared knowledge base to construct a responsible and ethical AI ecosystem.

ScaDS.AI Dresden/Leipzig is a leading German competence center dedicated to research in Artificial Intelligence, Big Data, and Data Science. Funded by the federal government's AI strategy, it operates as a permanent research facility with strong ties to TUD Dresden University of Technology and Leipzig University. The center expands upon the former ScaDS Dresden/Leipzig, combining expertise to bridge the gap between efficient mass data utilization, knowledge management, and advanced AI. It engages an international team of over 60 Principal Investigators and more than 180 employees, fostering interdisciplinary research across various domains including AI algorithms, applied AI, Big Data analytics, and responsible AI architectures. ScaDS.AI also provides extensive educational programs, training, software services, and consulting for professionals, students, and the general public through its Living Lab.

AgentLaboratory is an end-to-end autonomous research workflow tool designed to assist human researchers in implementing their research ideas. It leverages specialized large language model (LLM) agents to support the entire research process, from conducting literature reviews and formulating experimental plans to executing experiments and generating comprehensive reports. The system aims to complement human creativity by automating repetitive and time-intensive tasks like coding and documentation, allowing researchers to focus on ideation and critical thinking. It integrates external tools such as arXiv, Hugging Face, Python, and LaTeX, and supports various LLM models including OpenAI and DeepSeek. AgentLaboratory also features AgentRxiv, a framework for agents to collaboratively build on each other's research.

The Artificial Intelligence Research and Innovation Center (AIRIC) at Politecnico di Milano is dedicated to advancing AI research and fostering innovation. It serves as a crucial link between academic scientific research and the practical demands of industry and public administration. AIRIC actively engages in collaborative projects, developing innovative AI solutions that address real-world challenges. The center's work spans various aspects of AI, contributing to both theoretical advancements and their application in diverse sectors. This initiative highlights Politecnico di Milano's long-standing commitment to being at the forefront of AI research for over 50 years.

MiroFlow is a high-performance, modular, and fully open-source framework designed for building intelligent AI agents. It excels in multi-step internet research and complex challenges like future event prediction, achieving top rankings on multiple benchmarks including FutureX, GAIA, HLE, xBench-DeepSearch, and BrowserComp. The framework features advanced multi-turn conversation capabilities, extensive tool ecosystem integration, and hierarchical sub-agent orchestration. Built with robust concurrency management and fault-tolerant design, MiroFlow efficiently handles rate-limited APIs and unstable networks, ensuring reliable execution of complex tasks. It is cost-effective, powered by the open-source MiroThinker model, and can run a research agent service on a single RTX 4090, relying on free, open-source tools for simple deployment and scaling.

OAT (Online Alignment Toolkit) is a simple yet efficient open-source framework designed for running online LLM alignment algorithms. It features a distributed Actor-Learner-Oracle architecture optimized for high efficiency, utilizing vLLM for accelerated response sampling and DeepSpeed ZeRO for memory efficiency. OAT simplifies the experimental pipeline by providing an online Oracle for preference data labeling and real-time model evaluation. Researchers can simulate various feedback types, including verifiable rewards and LLM-as-a-judge, with flexible deployment options for reward models. Its modular structure facilitates rapid prototyping and experimentation, implementing cutting-edge algorithms like PPO/Dr.GRPO for online RL and Online DPO/SimPO/IPO for preference learning, fostering innovation and fair benchmarking.

Juristo AI is an advanced legal associate platform designed for high-stakes litigation, offering AI-powered legal intelligence and enterprise AI solutions. It automates legal research, drafting, and analysis using cognitive AI agents, ensuring citation accuracy and SOC2 compliance. The platform supports 7 languages and boasts 99.9% uptime. Juristo caters to law students with an AI tutor and research assistant, legal professionals with precision drafting and high-velocity research, and firms/enterprises with proprietary AI infrastructure for large-scale legal teams. Key features include intelligent document analysis, automated legal contract drafting, and structured conversational workflows, all built on a secure, proprietary AI core.

npcpy is a comprehensive Python library designed for research and development in NLP, multimodal LLMs, Agents, ML, and Knowledge Graphs. It offers a flexible agent framework that supports both local and cloud providers, enabling users to build sophisticated AI applications. Key capabilities include multi-agent team orchestration, tool calling, and advanced functionalities like image, audio, and video generation. The library also facilitates knowledge graph integration and fine-tuning of models, making it a versatile solution for developers and researchers working with diverse AI technologies. It provides quick examples for creating personas, direct LLM calls, agent with tools, streaming, JSON output, and Pydantic structured output.

ScholarAI is an all-in-one AI research assistant designed to revolutionize the research process for students, developers, and clinical teams. It allows users to search millions of peer-reviewed papers, summarize academic content, and generate citations in over 2,600 styles. The platform integrates with Jenni AI for drafting and auto-citing, and users can import research from Zotero or Mendeley with a single click. ScholarAI also offers an AI chat assistant, Agentic AI Chat, which consults the latest research and user-uploaded PDFs. For developers, ScholarAI provides an API to integrate its HIPAA-compliant research tools into custom applications.

The AgAID Institute is dedicated to building and nurturing collaborations between the artificial intelligence and agriculture sectors. Its core mission is to establish a transdisciplinary ecosystem that drives technological innovation and facilitates knowledge transfer within the agricultural domain. The institute focuses on various research areas including Artificial Intelligence Demo Farms, Labor Intelligence, Water Intelligence, and Farm Operations Intelligence. It also offers educational opportunities such as student internships, workforce training, and online courses on communicating AI in agriculture, alongside Digital AgATH0NS. The institute aims to address the evolving needs of agriculture through advanced AI applications and robust educational programs.

ResearchGOAT is a generative AI-powered platform designed for in-depth qualitative user research and interviews. It empowers clients in research, UX, and CX to automate the entire research process, from project planning to participant recruitment and AI-moderated interviews. The platform's AI conducts extensive live voice-to-voice video recorded interviews, intelligently formulating follow-up questions based on customer responses. This allows for boundless research, unrestricted by time zones or geographical locations, and provides dynamic insights into customer worlds anytime, anywhere, and in any language. ResearchGOAT dramatically reduces costs and delivers insights faster than traditional methods, making research adaptive, cost-effective, and infinitely scalable.

The Institute for Law & AI (LawAI) is an independent think tank dedicated to researching and advising on the complex legal challenges presented by artificial intelligence. LawAI conducts foundational research into the intersection of AI and law, aiming to provide governments, industry, academics, and the public with a deeper understanding of AI governance. Beyond research, LawAI offers consulting services to governments, international organizations, and private actors, helping them navigate the legal landscape of AI. The institute also actively publishes its findings through research articles, books, and blog posts, contributing to the ongoing discourse on AI's societal impact. They are committed to fostering sound legal analysis to ensure AI development aligns with security, welfare, and the rule of law.

Gatsbi is an AI-powered research assistant designed to streamline the innovation process for researchers, engineers, students, and enterprises. It helps users discover original research ideas by detecting research gaps and generating innovative solutions with originality scores and relevant references. The platform can auto-draft academic papers with citations, figures, and equations based on research notes or prior materials, supporting methodological, experimental, and case-study research. Gatsbi also automates systematic literature reviews and meta-analyses, including study screening, data extraction, and statistical synthesis. Additionally, it can draft comprehensive patent disclosure documents in multiple languages, covering all key sections from description to claims. Gatsbi offers both web and desktop versions (Windows, macOS) and integrates a Humanizer plugin to reduce the likelihood of AI detection.

Wikipedia2Vec is a powerful tool for obtaining vector representations (embeddings) of words and entities directly from Wikipedia. Developed and maintained by Studio Ousia, it allows users to learn embeddings for concepts that have corresponding Wikipedia pages. The tool implements the conventional skip-gram model for word embeddings and an extension for entity embeddings, enabling similar words and entities to be positioned closely in a continuous vector space. Embeddings can be easily trained using a single command with a publicly available Wikipedia dump as input. Pretrained embeddings are available for 12 languages, and the tool has been applied to various NLP tasks including entity linking, named entity recognition, question answering, and text classification.

ArXiv Pulse is an AI-powered platform designed to help researchers stay informed about the latest academic papers without being overwhelmed. It delivers clear, easy-to-read summaries of arXiv preprints directly to your inbox, tailored to your specific research interests. Users can customize their research preferences and choose how often they receive updates, ensuring relevant content is delivered consistently. The platform aims to ease the research update load, helping users avoid missing crucial developments that could impact their work or inspire new projects. It offers both topic-specific and general field overviews, powered by advanced AI models like Gemini 1.5 Flash and Pro.

The Kingston AI Group is a collective of Australian AI professors dedicated to shaping the future of artificial intelligence in Australia. The group emphasizes the need for a national vision and strategy for AI, advocating for substantial new investments to build core sovereign capabilities. This initiative is crucial for Australia's prosperity, economic diversity, and sovereignty in the rapidly transforming global economy. The group's members are experts from various fields, collaborating to advance AI research and education across the nation.

doubleAI focuses on advancing machine learning to achieve next-level AI, specifically aiming to solve complex challenges in computer science, mathematics, and science. The company's research is dedicated to developing superintelligence and improving reasoning capabilities within AI systems. A key offering is WarpSpeed, a technology designed to surpass expert-written kernels, making NVIDIA's best GPU code even faster. This innovation is geared towards scaling expert intelligence and progressing humanity through advanced AI applications, particularly in scientific and technological domains. The tool emphasizes research and development to push the boundaries of AI performance.

LilysAI is an AI-native content distilling assistant designed to accelerate knowledge acquisition and research. Users can input various materials, including YouTube videos, audio files, PDFs, websites, and text, to receive instant summaries, clear explanations, and deeper insights. The tool allows for saving sources, subscribing to channels and RSS feeds, and generating structured reports with customizable length and depth. Beyond summarization, LilysAI offers features like auto-generated reports with background knowledge, opposing viewpoints, actionable items, mind maps, infographics, quizzes, and flashcards. It also provides in-depth explanations for complex ideas and an AI chat agent that remembers user interests and proactively starts conversations. LilysAI helps users discover and recommend valuable sources based on their interests, making it an indispensable tool for efficient information intake.

Outset is an AI-moderated research platform designed to provide the depth of qualitative interviews with the speed and scale of a survey. It leverages conversational AI, behavioral intelligence, and emotional analysis to bridge the gap between what consumers say and do. The platform offers a comprehensive workflow from AI-guided interview setup and participant recruitment to AI-moderated interviews in multiple languages (video, voice, text). Key features include instant synthesis and analysis, generating summaries, themes, powerful quotes, and customizable highlight reels. Outset also facilitates easy stakeholder shareout with AI-generated reports and presentations, making research a core part of development and strategy. It supports various use cases like discovery, evaluative, market strategy, personas, brand research, concept testing, UX & usability, UX evals, and creative testing.