The Reference Set Is Not the Answer Set: GEOM-DRUGS Is Missing Low-Energy Conformers on Its Own Energy Surface
Abstract
Machine-learned conformer generators are scored on GEOM-DRUGS by proximity to a reference ensemble that standard benchmark metrics implicitly treat as ground truth. We test whether that finite set is complete on the GFN2-xTB surface used to construct it. Across five public generators, settled, low-energy structures remain outside the $0.75Å reference radius after removing reflections, incorrect stereochemistry, and post−relaxation connectivity changes (2.3−−7.0\%$ of submitted conformers, depending on architecture). In a targeted DFT audit, $107$ of $113$ chemically screened novel/reference pairs remain outside that radius after full ωB97X-D3BJ re-optimization. A threshold-free complementary test needs no radius and no energy window at all: on 15.0% $[8.0, 22.0]$ of molecules some generator has already found a chemically intact structure lying below the \emph{entire} released ensemble. A complementary CREST search reproduces the minimum released in $65$ of $69$ molecules, and in $11$ of them it reaches more than $0.5$\,kcal/mol below the complete released ensemble. Augmenting the references with conformers proposed by held-out architectures raises absolute coverage-precision for every generator tested, by $1.48−−2.99$ points. Together, these results show that GEOM-DRUGS precision is sensitive not only to generator quality but also to incomplete reference support.
Related
- Analytic Energy Gradients for Unrestricted Mixed-Reference Spin-Flip Time-Dependent Density Functional Theory
- Spectroscopic identification of gauche and trans conformers in L-phenylalanine Fiber and hydrogel via CH/CD stretching vibrations
- A basis-set-free reference for the exact exchange-energy density and functional development
- A single reference-energy outlier nearly eliminates Co–O chemistry from a generative model's training distribution
- Benchmarking a fast energy model for predicting the stability of solvated surface sites for organic molecular crystals: The paracetamol–water interface
BibTeX
@misc{vahediahmar2026reference,
title = {{The Reference Set Is Not the Answer Set: GEOM-DRUGS Is Missing Low-Energy Conformers on Its Own Energy Surface}},
author = {Alvand Vahediahmar and Aaron Zweig and Mingxuan Zhang and Elham Azizi},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.12523},
eprint = {2610.12523},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.12523},
note = {CC BY-NC-SA 4.0},
}
RIS
TY - UNPB
TI - The Reference Set Is Not the Answer Set: GEOM-DRUGS Is Missing Low-Energy Conformers on Its Own Energy Surface
AU - Alvand Vahediahmar
AU - Aaron Zweig
AU - Mingxuan Zhang
AU - Elham Azizi
PY - 2026
DA - 2026/10/08
PB - arXiv
DO - 10.48550/arxiv.2610.12523
UR - https://arxiv.org/abs/2610.12523
AB - Machine-learned conformer generators are scored on GEOM-DRUGS by proximity to a reference ensemble that standard benchmark metrics implicitly treat as ground truth. We test whether that finite set is complete on the GFN2-xTB surface used to construct it. Across five public generators, settled, low-energy structures remain outside the $0.75Å reference radius after removing reflections, incorrect stereochemistry, and post−relaxation connectivity changes (2.3−−7.0\%$ of submitted conformers, depending on architecture). In a targeted DFT audit, $107$ of $113$ chemically screened novel/reference pairs remain outside that radius after full ωB97X-D3BJ re-optimization. A threshold-free complementary test needs no radius and no energy window at all: on 15.0% $[8.0, 22.0]$ of molecules some generator has already found a chemically intact structure lying below the \emph{entire} released ensemble. A complementary CREST search reproduces the minimum released in $65$ of $69$ molecules, and in $11$ of them it reaches more than $0.5$\,kcal/mol below the complete released ensemble. Augmenting the references with conformers proposed by held-out architectures raises absolute coverage-precision for every generator tested, by $1.48−−2.99$ points. Together, these results show that GEOM-DRUGS precision is sensitive not only to generator quality but also to incomplete reference support.
ER -