SOURCE-LINKED INTELLIGENCE

Inherited Heads: Audio language models track speakers with their text backbone's attention, and an attention-mass ranking retrieves a different set

arXiv · AI, language, vision and robotics · article · Sep 12, 2026 · UTC

Asked to describe what one of six speakers in a recording talks about, audio language models describe the right one on 6 to 16% of trials, below the 16.7% a guess would give. Adding a fixed bias to the attention logits of a hundred heads, under a tenth of the model's and with no training, redirects the description to whichever speaker we choose, on 90.7% to 99.0% of trials. Those heads are largely not specific to audio. Rank the text-only language model an audio model was built from, or a released model of the same family, on a written version of the task, take its top hundred heads, and carry

Read original source ↗ Open in workspace

recordType

paper

region

Global