Nicolas Thome

Synthetic scenes with automatically grounded object identities unlock substantial real-world perceptual gains in MLLMs via on-policy self-distillation.

TSPFN outperforms both traditional tabular models and specialized deep learning approaches in classifying physiological time series, showcasing its superior generalization capabilities.

MODIP outperforms traditional behavioral cloning and rivals advanced reinforcement learning methods, showcasing a new frontier in efficient policy fine-tuning for robot learning.

MLLMs can get a surprising visual reasoning boost from a simple trick: adding just a dash of visually grounded self-supervision to instruction tuning.