On Representation Learning-based Methods for Effective, Efficient, and Scalable Code Retrieval

Celso França, Rennan C. Lima, Cláudio M. V. de Andrade, Washington Cunha, Pedro O. S. Vaz de Melo, Berthier A. Ribeiro-Neto, Leonardo Rocha, Rodrygo L. T. Santos, Adriana S. Pagano, Marcos André Gonçalves

Abstract: Highlights•xCoFormer fully meets effectiveness, efficiency, and scalability requirements.•Tag-training, a two-player turn-based bargaining game method, improves training efficiency.•The xCoFormer’s attention mechanism provides hints about the code retrieval process.•xCoFormer yields the best cost-effectiveness trade-off in the code retrieval task.