Youssef Mroueh

Karthikeyan Shanmugam

Peter Hess

Pin-Yu Chen

Pierre Dognin

Payel Das

Aleksandra Mojsilovic

Samuel Hoffman

 Vijil Chenthamarakshan

Kahini Wadhawan

Kim Martineau

News

Research

Deep Dive

Explainer

What is retrieval-augmented generation?

Teaching AI models to improve themselves 

 DualTKB: A Dual Learning Bridge between Text and Knowledge Base

 Image captioning as an assistive technology

AI boosts the discovery of metamaterials vital for next-gen gadgets

Accelerating molecular optimization with AI

IBM AI finds new peptides – paving the way to better drug design

ICASSP 2025

IWCT 2025

ICLR 2025

NAACL 2025

Trustworthy AI

Trustworthy Generation

Computer Vision

Knowledge and Reasoning

Natural Language Processing

Generative AI

Computer Science

Materials Discovery

Healthcare

Explainable AI

Accelerated Discovery

Uncertainty Quantification

Data is key to technological innovations. We develop theoretical and algorithmic frameworks for generative AI to synthesize realistic, diverse, and targeted data. Our methods facilitate data augmentation for trustworthy machine learning and accelerate novel designs for drug and material discovery, and beyond.

Machine learning uses data to teach AI systems to imitate the way that humans learn. They can find the signal in the noise of big data, helping businesses improve their operations. We've been in the field since since the beginning: IBMer Arthur Samuel even coined the term “Machine Learning” back in 1959. 

Machine Learning

It can take over 10 years to come up with new materials. At IBM Research, we’re looking to accelerate the discovery process using new AI methods, robotics, the hybrid cloud, and quantum computers. Our goal is to unlock new properties and materials to address global challenges in years not decades.

Much of the information that can help transform enterprises is locked away in text, like documents, tables, and charts. We’re building advanced AI systems that can parse vast bodies of text to help unlock that data, but also ones flexible enough to be applied to any language problem.

Biases can lead to systematic disadvantages for marginalized individuals and groups — and they can arise in any point in the AI development lifecycle. To increase the accountability of high-risk AI systems, we're developing technologies to increase their end-to-end transparency and fairness.

Fairness, Accountability, Transparency

We’re developing technologies to increase the end-to-end transparency and fairness of AI systems. 

Animate gif with that transitions among of images derived using a generative adversarial network (GAN) trained on images of viruses and molecules.

Example of a metasurface without light direction.

IBM Research scientists are working on different approaches to teach AI models how to improve themselves

Trustworthy Generation

Our work

Teaching AI models to improve themselves

What is retrieval-augmented generation?

Accelerating molecular optimization with AI

AI boosts the discovery of metamaterials vital for next-gen gadgets

IBM AI finds new peptides – paving the way to better drug design

DualTKB: A Dual Learning Bridge between Text and Knowledge Base

Image captioning as an assistive technology

Publications

The Literary Canons of Large-Language Models: An Exploration of the Frequency of Novel and Author Generations Across Gender, Race and Ethnicity, and Nationality

A transfer learning framework for weak to strong generalization

Out-of-Distribution Detection using Synthetic Data Generation

Large Language Models can Become Strong Self-Detoxifiers

Contextual Value Alignment

Combinatorial Test Design Model Creation using Large Language Models

Related topics