Research Tools: code, datasets, & models
Discover an index of datasets, SDKs, APIs and open-source tools developed by Microsoft researchers and shared with the global academic community below. These experimental technologies—available through Azure AI Foundry Labs (opens in new tab)—offer a glimpse into the future of AI innovation.
CherIoT-ibex: reference implementation of the ISA
cheriot-ibex is a RTL implementation of CHERI ISA based on LowRISC’s Ibex core. More specifically, cheri-ibex is a 32-bit RISC-V microcontroller which implements the CheriIoT ISA extension in addition to RV32IMCB. Same as the original…
CherIoT Sail: formal specification of the ISA
This repository contains an implementation of the CHERIoT ISA in Sail. It contains an executable description of the CHERIoT instruction set that can be used to build an instruction set emulator and also prove properties…
CherIoT RTOS and software stack
This repository contains the core RTOS components for the CHERIoT research platform. This is currently a research project that has been open-sourced to enable wider collaboration. It is not yet in a state where it…
SPADE: State Space Augmented Transformer
This PyTorch package implements the language modeling experiments in Efficient Long Sequence Modeling via State Space Augmented Transformer.
DeFacto: Natural Language Generation from Natural Interactions
DeFacto is a dataset containing human demonstrations and feedback for improving the factual consistency of text summarization.
CityLifeSim: A High-Fidelity Pedestrian and Vehicle Simulation with Complex Behaviors
CityLife is a flexible, high-fidelity simulation that allows users to define complex scenarios with essentially unlimited actors, including both pedestrians and vehicles. This tool allows each vehicle and pedestrian to operate with basic intelligence that…
TorchScale – A Library for Transformers at (Any) Scale
TorchScale is a PyTorch library that allows researchers and developers to scale up Transformers efficiently and effectively. It has the implementation of fundamental research to improve modeling generality and capability as well as training stability…
Motion Capture with Actions (MoCapAct)
The MoCapAct dataset contains training data and models for humanoid locomotion research. It consists of expert policies that are trained to track individual clip snippets and HDF5 files of noisy rollouts collected from each expert,…
Orchestrated Value Mapping
This repository hosts the code release for the paper “Orchestrated Value Mapping for Reinforcement Learning”, published at ICLR 2022. This work was done by Mehdi Fatemi (Microsoft Research) and Arash Tavakoli (Max Planck Institute for…