Contributed Talk Sessions | Poster Sessions | All Posters | Search Papers

Poster Session B: Wednesday, August 13, 1:00 – 4:00 pm, de Brug & E‑Hall

Decoding Signatures of Meta-Learning Abstract Structure

Doris Yu¹, Sreejan Kumar², Marcelo G Mattar³; ¹University of Washington, ²Princeton University, ³New York University

Presenter: Sreejan Kumar

Humans excel at learning abstract structure from limited data and applying it to novel situations—a capacity often attributed to meta-learning. While behavioral evidence supports this ability, the neural mechanisms by which abstract concepts are acquired and refined during learning remain unclear. In this study, we use magnetoencephalography (MEG) to examine how the brain dynamically constructs abstractions while learning a set of tasks generated with a compositional grammar. Through MEG decoding, our results show evidence of learning the grammar structure across multiple timescales, both within and across different trials. These findings provide neural evidence for meta-learning in humans, showing that abstract representations emerge during learning.

Topic Area: Memory, Spatial Cognition & Skill Learning

Extended Abstract: Full Text PDF