Can you explain the trade-offs between different architectures for sequence modeling for Docusign document workflows?
