A 360 review of AI agent benchmarks
Research
Kim Martineau
Trustworthy Generation
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.
Our work
An invisible watermark to keep tabs on tabular data
ResearchKim MartineauTeaching AI models to improve themselves
ResearchPeter HessWhat is retrieval-augmented generation?
ExplainerKim MartineauAccelerating molecular optimization with AI
Deep DivePayel Das, Samuel Hoffman, Vijil Chenthamarakshan, Kahini Wadhawan, and Pin-Yu Chen11 minute readAI boosts the discovery of metamaterials vital for next-gen gadgets
ResearchYoussef Mroueh, Karthikeyan Shanmugam, and Payel Das10 minute read- See more of our work on Trustworthy Generation
Publications
Position: Theory of Mind Benchmarks are Broken for Large Language Models
- Matthew Riemer
- Zahra Ashktorab
- et al.
- 2025
- ICML 2025
The Literary Canons of Large-Language Models: An Exploration of the Frequency of Novel and Author Generations Across Gender, Race and Ethnicity, and Nationality
- Paulina Toro Isaza
- Nalani Kopp
- 2025
- NAACL 2025
Exploring Straightforward Methods for Automatic Conversational Red-Teaming
- George Kour
- Naama Zwerdling
- et al.
- 2025
- NAACL 2025
A transfer learning framework for weak to strong generalization
- Seamus Somerstep
- Felipe Maia Polo
- et al.
- 2025
- ICLR 2025
Out-of-Distribution Detection using Synthetic Data Generation
- Momin Abbas
- Muneeza Azmat
- et al.
- 2025
- ICLR 2025
Large Language Models can Become Strong Self-Detoxifiers
- Irene Ko
- Pin-Yu Chen
- et al.
- 2025
- ICLR 2025