Renpu Liu

Renpu Liu

Office: Room 328 Rice Hall

On the Learn-to-Optimize Capabilities of Transformers in In-Context Sparse Recovery

Published in ICLR 2025, 2025

We demonstrate that Transformers can implement LISTA-type learning-to-optimize algorithms for LASSO-based sparse recovery.

Recommended citation: Renpu Liu*, Ruida Zhou*, Cong Shen, and Jing Yang. (2025). "On the Learn-to-Optimize Capabilities of Transformers in In-Context Sparse Recovery." ICLR 2025.
Download Paper