ChemBO Baselines

This repository packages the baseline methods and helper scripts extracted from the local ChemBO-Agent workspace into a standalone release layout.

Included:

• baseline/BO-ICL
• baseline/Pref-BO
• baseline/Reasoning-BO
• baseline/bora
• baseline/gollum
• baseline/common/tabular_benchmarks.py
• run_tabular_baseline.py as a unified experiment launcher
• scripts/ with baseline-focused run scripts
• data/DAR.csv, data/OCM.csv, and data/suzuki.csv for the tabular benchmark runs

Not included:

• ChemBO Agent main optimization code
• experiment outputs
• local caches and nested git metadata

Notes:

• Some baseline subprojects preserve their own upstream layout and dependency files.
• The run scripts assume this repository root contains baseline/, scripts/, and data/.
• The unified launcher can forward dataset-specific and baseline-specific CLI arguments while writing each run into its own timestamped output directory.
• scripts/setup_baseline_envs.sh bootstraps the per-baseline conda environments that are used in this repo.
• scripts/run_suzuki_reasoningbo_gollum_3x.sh runs the requested Suzuki benchmark suite for Reasoning-BO and GOLLuM.