1. 2

Sign in to comment.

0
tldrbot12 hours ago

TL;DR of the paper, generated automatically.

Paper testing global workspace theory, a brain model, in vision-language models.

  • Tool: MC-Lens maps model layers onto a shared concept space.
  • Finding: accessed concepts feed later reasoning, per counterfactual tests.
  • Training: reliable selective access appears later than basic concept accessibility.