Anthropic introduces Jacobian Lens to map a cognitive-like global workspace representation inside transformer models
The interpretability method computes a single matrix per layer.
Story Brief
The interpretability method computes a single matrix per layer.
Related Stories
3 linkedVerbalizable Representations Form a Global Workspace in Language ModelsIf the mind is an ocean, we spend our lives floating at the surface. Beneath us, an enormous amount of processing takes place without our knowledge: our visual systems parsing the contours of a face, our motor circuits maintaining our posture. At any given moment, only a small fraction of this neural activity is accessible to us. Yet it is this privileged sliver of activity that we rely on to reason deliberately: to plan what ingredients to buy for a recipe, or to puzzle out why an engine won’t start. Such thoughts can be articulated out loud, deliberately held in mind, and brought to bear on whatever task the moment demands. This distinction, between our accessible thoughts and our unconscious processing, is perhaps the most striking feature of human cognition.13h agotransformer-circuits.pub
A global workspace in language modelsInterpretability research on Claude's internal thoughts.13h agoAnthropicAInature.comSource context from nature.com.13h agonature.com
A global workspace in language modelsInterpretability research on Claude's internal thoughts.13h agoAnthropicAInature.comSource context from nature.com.13h agonature.com
