kellogh, 1 year ago Yesterday I saw this, which is similar. But I don’t think what they’re doing is as simple as “reading the logits”. Seems like they had to train a model to interpret them. So I guess it’s not out of the question https://syncedreview.com/2022/11/16/deepminds-epistemic-neural-networks-enable-large-language-model-fine-tuning-with-50-less-data/
Yesterday I saw this, which is similar. But I don’t think what they’re doing is as simple as “reading the logits”. Seems like they had to train a model to interpret them. So I guess it’s not out of the question https://syncedreview.com/2022/11/16/deepminds-epistemic-neural-networks-enable-large-language-model-fine-tuning-with-50-less-data/