Research Tools: code, datasets, & models

Prototype experimenting with agents in a grid UI — simple, open experiment that illustrates how agent workflows might live in a grid-like surface.

Magentic Marketplace is an open-source simulation environment for exploring the numerous possibilities of agentic markets and their societal implications at scale. It provides a foundation for studying these markets and guiding them toward outcomes that benefit everyone.

MMCTAgent (Multi-modal Critical Thinking Agent) is a state-of-the-art multi-modal AI framework that brings human-like critical thinking to visual reasoning tasks. it combines advanced planning, self-critique, and tool-based reasoning to deliver superior performance in complex image…

Open Source fuels the AI for Good Lab’s mission by providing research tools, data, and code that drive innovation, foster collaboration, and enhance accessibility for the global community. Explore our full library of resources to…

MCP Interviewer is a Python CLI tool that helps you catch MCP server issues before your agents do. The MCP Interviewer ensures compliance with provider constraints and offers warnings when recommended guidance is not followed, while…

RetroChimera is a model that takes as input a target molecule that one wants to synthesize, encoded as a sequence of characters (using the SMILES notation), and produces several potential chemical reactions which could be…

Codec for converting binary strings into DNA-like sequences over the alphabet {A, C, G, T} while guaranteeing that no symbol repeats more than k times consecutively (length k homopolymer run). The repository ships a fast…

Skala is a neural network-based exchange-correlation functional for density functional theory (DFT), developed by Microsoft Research AI for Science. It leverages deep learning to predict exchange-correlation energies from electron density features, achieving chemical accuracy for…

This repository hosts the official code and data artifact for the “Affective Air Quality” dataset. Details about the user study and data collection can be found in our paper (opens in new tab). The dataset released…

Reinforcement learning (RL) finetuning method that ensures that the inference-time compute for queries is optimized based on query difficulty, leading to significant inference efficiency.