This visualization maps the optimization workflow for the PXR ligand-pose prediction challenge, showing how automated method variants are iteratively tested against a leakage-controlled LDDT-PLI benchmark. It charts performance scores for co-folding models, diffusion guidance, docking hybrids, and fine-tuning strategies, with each bar or point colored by method family. The code uses `d3.forceSimulation` for node layout and `d3.scaleLinear` for axis mapping, drawing from the gist’s computational framework description and benchmark results. The chart highlights that IntelliFold-2 and physics-guided steering outperform the Boltz-2 baseline, while most sampling and refinement approaches yield flat or negative gains.