Scaling can lead to compositional generalization

NeurIPSSpotlight2025

Authors
Florian Redhardt, Yassir Akram, Simon Schug
Venue
NeurIPS 2025
Track
Spotlight

TL;DR

Can neural networks systematically capture discrete, compositional task structure despite their continuous, distributed nature? The impressive capabilities of large-scale neural networks suggest that…

Opening excerpt from the authors’ abstract. source

Read the paper

Topics

generalization

← All NeurIPS 2025 Spotlight papers · Browse the whole archive