Oct 132026 Like humans, language models demonstrate face-to-character biases Abstract. Humans routinely make unjustified character inferences, such as labeling people as trustworthy or untrustworthy, based on facial features. We con
