Microsoft

Channel-adaptive autoregressive generative model that synthesizes in-silico fluorescence microscopy of protein subcellular localization from amino-acid sequence and cellular landmark stains.

A language model that generates small-molecule structures directly from transcriptomic phenotypes — gene up/down-regulation signatures — for phenotype-driven drug discovery.

Lightweight 7B vision-language foundation model from Microsoft Research, released research-only under the Microsoft Research License, that generates radiology findings from chest X-rays.

Unified science foundation model from Microsoft Research treating molecules, proteins, RNA, DNA, and materials as a shared sequence language for cross-domain generation.

Generative deep learning model from Microsoft Research that emulates protein equilibrium ensembles at 100,000x the speed of molecular dynamics simulation.

A biomedical foundation model for joint segmentation, detection, and recognition across nine imaging modalities using natural language prompts.

A transformer protein language model using integrative co-evolutionary pre-training to capture both short-range and long-range residue interactions from sequence alone.

A multi-task EEG foundation model that treats brain signals as a foreign language, pairing a text-aligned neural tokenizer with a GPT-2 backbone.

An enhanced T5-based encoder-decoder that unifies molecule, protein, and text understanding via IUPAC integration and multi-task instruction tuning.

Lightweight variant of AlphaFlow achieving ~47x faster conformational ensemble sampling by fine-tuning only AlphaFold's structure module with frozen Evoformer.

Microsoft Research multimodal LLM for grounded chest X-ray report generation, localizing each described finding with bounding boxes on the image.

Whole-slide pathology foundation model pretrained on 1.3 billion tiles from 171,189 clinical WSIs. Achieves state-of-the-art on 25 of 26 pathology benchmark tasks.

Deep learning framework predicting equilibrium distributions of molecular systems, enabling efficient ensemble generation and conformation sampling.

A transformer EEG foundation model pretrained with vector-quantized self-supervision on 1.7 TB of EEG, yielding transferable, interpretable discrete representations.

A masked-autoencoder EEG pretraining framework that maps any electrode layout to a unified topology for topology-agnostic, cross-dataset representations.

Microsoft Research multimodal LLM that generates the findings section of a chest X-ray report from a single frontal image using a CXR-specific vision encoder and Vicuna-7B.

Sequence-first protein generation framework using discrete diffusion over evolutionary alignments, enabling controllable de novo design without structure.

Graph neural network framework for antigen-specific antibody CDR design, combining a pre-trained antibody language model with one-shot sequence and structure generation.

A biomedical vision-language assistant from Microsoft Research, adapted from LLaVA via curriculum learning on PubMed Central figure-caption pairs and GPT-4-generated instructions.

Multimodal biomedical foundation model trained on 15M PubMed Central figure-caption pairs via contrastive learning, achieving state-of-the-art zero-shot performance across imaging modalities.

A GPT-2-based generative transformer pretrained on 15M PubMed abstracts for biomedical text generation and mining, including relation extraction and question answering.

CNN-based protein language model series showing convolutions match transformer performance on sequence pretraining while scaling linearly with sequence length.