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.
PIKE-RAG (sPecIalized KnowledgE and Rationale Augmented Generation) framework mainly consists of several basic modules, including document parsing, knowledge extraction, knowledge storage, knowledge retrieval, knowledge organization, knowledge-centric reasoning, and task decomposition and coordination while building coherent…
PadChest-GR is a manually annotated, bilingual chest X-ray dataset designed to train and evaluate models for grounded radiology report generation. It includes bounding boxes and comprehensive annotations of all clinically relevant findings.
Text Adventure Learning Environment Suite (TALES) – Benchmark to evaluate language models on interactive text environments. This repository contains the files needed to benchmark language agents on a curated list of text-based games from the…
Implementation of Discrete Attn Transformer. Includes from scratch training of transformer for Templatic Generation Tasks (dataset available at: https://huggingface.co/datasets/rfernand/templatic_generation_tasks), along with code to evaluate LLMs on same dataset. See our paper: “Mechanisms of Symbol Processing…
Scalable long-context LLM decoding that leverages sparsity—by treating the KV cache as a vector storage system.
Structured Entity Extraction and the Approximate Entity Set OverlaP (AESOP) metric are designed to appropriately assess model performance.
AttentionEngine accelerates transformer attention variants by generating efficient custom kernels, enabling model designers to easily create new variants with our flexible API.
Understanding land use over time is critical to tracking events related to climate change, like deforestation. However, satellite-based remote sensing tools which are used for monitoring struggle to differentiate vegetation types in farms and orchards…
SeerAttention is a learning-based method to enable block-level sparse attention for long-context LLM without using prefined static pattern or heuristic methods. It can be applied in Post-training or Fine-tuning stages. The Attention Gate units learn…
