Jonathan D. Cohen

Bound by semanticity: universal laws governing the generalization-identification tradeoff

Visual serial processing deficits explain divergences in human and VLM reasoning

Binding Visual Features Point by Point

Learning is Forgetting; LLM Training As Lossy Compression

Emergent Symbolic Mechanisms Support Abstract Reasoning in Large Language Models

Causal Head Gating: A Framework for Interpreting Roles of Attention Heads in Transformers