Robust and Efficient Finetuning of Vision Foundation Models via Implicit Ensembling

Name: Robust and Efficient Finetuning of Vision Foundation Models via Implicit Ensembling
Uploaded: 2026-10-20T20:18:56.881850+00:00
Duration: 460 s
Description: GMixout extends Mixout to improve robustness when finetuning vision foundation models under distribution shifts. It introduces an adaptive exponential moving average anchor and a resampling frequency hyperparameter. This method enhances out-of-distribution generalization while maintaining strong in-domain accuracy and achieving high efficiency on consumer GPUs.

GMixout extends Mixout to improve robustness when finetuning vision foundation models under distribution shifts. It introduces an adaptive exponential moving average anchor and a resampling frequency hyperparameter. This method enhances out-of-distribution generalization while maintaining strong in-domain accuracy and achieving high efficiency on consumer GPUs.

Open the full episode