OptFormer Project: Research in Language Modeling for Optimization

Codebase Organization

This codebase is organized by general-purpose library folders and specific projects:

General-Purpose Libraries

• Common Imports
• T5X Modules
• PAX Modules

Optimization API Wrappers

• Vizier
• PyGlove

Specific Papers and Projects

• Original OptFormer (Chen et al., 2022)
• OmniPred (Song et al., 2024)
• Understanding LLM Embeddings (Tang et al., 2024)
• Embed-then-Regress (Nguyen et al., 2024)
• Decoding-based Regression (Song and Bahri, 2025)

Disclaimer: This is not an officially supported Google product.