Soumya Ghosh

New methods enhance reliability of AI predictions, particularly in critical fields.

A new method enhances probability accuracy in language models without extra data.

MExGen framework improves explanations for generative language models, enhancing user trust.

A new method enables the transfer of LoRA modules with synthetic data, minimizing reliance on original data.

A new model detects social bias in text using synthetic data.

Explore the significance and methods of Training Data Attribution in AI.