Visual similarity partly explains adult and model word extension
DINOv2 distance tracked adult extension, and an artificial learner was less accurate for farther category pairs.
Stories per day · 30 days
Relatedcognitive-science1dinov21interpretability1language-models1safety-readouts1sparse-autoencoders1visual-similarity1word-learning1
3 stories, newest first · one card per story
Stories tagged #representation
Redistribution score tracks recurring SAE feature shifts
On two frozen safety readouts, median supports of about 20 features reached failure AUROCs of 0.935 and 0.958.
Imprint Reader: From Weight-Update Readout to Behavioral Intervention
An Imprint Reader describes frozen weight updates and guides edits that raise refusal and tool-use scores.