CONTACT: [email protected]
This repository aims to benchmark several different LLM methods for Bayesian Optimization across two domains.
Bayesian Optimization of protein motifs with LLMs. Benchmark traditional Bayesian Optimization methods against LLM-based approaches for optimizing 4-amino acid protein motifs. Includes generative reasoning models (GPT-5, Qwen3, DeepSeek-R1), agentic workflows, and fine-tuned models trained on synthetic optimization trajectories. See ALDE/README.md.
Molecular property prediction active learning framework. Modular benchmarking of LLM-based selectors vs traditional methods for molecular discovery, with multiple oracle models, selectors, and ablation tools. See mols/README.md.
Each subdirectory contains its own detailed README with installation instructions, usage examples, and technical documentation.
