Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Looped Transformers are Better at Learning Learning Algorithms

About

Transformers have demonstrated effectiveness in in-context solving data-fitting problems from various (latent) models, as reported by Garg et al. However, the absence of an inherent iterative structure in the transformer architecture presents a challenge in emulating the iterative algorithms, which are commonly employed in traditional machine learning methods. To address this, we propose the utilization of looped transformer architecture and its associated training methodology, with the aim of incorporating iterative characteristics into the transformer architectures. Experimental results suggest that the looped transformer achieves performance comparable to the standard transformer in solving various data-fitting problems, while utilizing less than 10% of the parameter count.

Liu Yang, Kangwook Lee, Robert Nowak, Dimitris Papailiopoulos• 2023

Related benchmarks

TaskDatasetResultRank
Sudoku SolvingSudoku-Extreme (test)
Accuracy61.3
31
Graph ColoringGraph Coloring 10-vertex
Conflict157.3
8
N-QueensN-Queens 10 x 10
Accuracy50
8
N-QueensN-Queens 8 x 8
Accuracy68.4
8
Graph ColoringGraph Coloring 8-vertex
Conflict136
8
Showing 5 of 5 rows

Other info

Follow for update