AI Frontiers
다운로드
Phi-4
6월 2025
Phi-4-multimodal and Phi-4-mini, the newest models in Microsoft’s Phi family of small language models (SLMs) are now available. These models are designed to empower developers with advanced AI capabilities. Phi-4-multimodal, with its ability to process speech, vision, and text simultaneously,…
Magentic-UI
4월 2025
Magentic-UI is a research prototype of an agentic web interface for solving complex web tasks. An Orchestrator coordinates four AutoGen agents—WebSurfer, Coder, FileSurfer, and UserProxy—to handle browsing, coding, file management, and user feedback, etc. It is designed with user-agent collaboration…
Steering LLMs for better instruction following
3월 2025
This repository contains the code for the paper “Improving Instruction-Following in Language Models through Activation Steering,” presented at ICLR 2025.
Skill Slice Insights
1월 2025
This is the official code repository for the paper “Unearthing Skill-level Insights for Understanding Tradeoffs of Foundation Models”. All rationales, localized skills, and skill-slices for the 12 datasets studied in the paper can also be accessed through this repo.
Magentic-One
11월 2024
Magentic-One is a generalist multi-agent system created to address intricate web and file-based tasks. By utilizing an intelligent Orchestrator alongside specialized agents, it facilitates the automation of complex, multi-step activities across various environments.
OmniParser V2
10월 2024
OmniParser is an advanced vision-based screen parsing module that converts user interface (UI) screenshots into structured elements, allowing agents to execute actions across various applications using visual data . By harnessing large vision-language model capabilities, OmniParser improves both efficiency and…
Trace
7월 2024
Trace is a new AutoDiff-like tool for training AI systems end-to-end with general feedback (like numerical rewards or losses, natural language text, compiler errors, etc.). Trace generalizes the back-propagation algorithm by capturing and propagating an AI system’s execution trace. Trace…
Phi-3
4월 2024
The Phi-3-Mini-128K-Instruct is a 3.8B parameters, lightweight, state-of-the-art open model trained with the Phi-3 datasets that includes both synthetic data and the filtered publicly available websites data with a focus on high-quality and reasoning dense properties. The model belongs to…
KITAB Dataset
2월 2024
🕮 KITAB is a challenging dataset and a dynamic data collection approach for testing abilities of Large Language Models (LLMs) in answering information retrieval queries with constraint filters. A filtering query with constraints can be of the form “List all books…