AI Paper Atlas

Research map / Probabilistic and scientific modeling / Molecules and proteins

Derivative Informed Learning of Exchange-Correlation Functionals

Eike Eberhard, Luca Thiede, Abdulrahman Aldossary, Andreas Burger, Nicholas Gao, Vignesh C Bhethanabotla, Alan Aspuru-Guzik, Stephan Günnemann

ICML 2026 · Probabilistic and scientific modeling › Molecules and proteins · approach: Machine learning for physical simulation · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add derivative-informed loss (gradient and Hessian supervision) to energy+density lossfour ML-XC architectures · B3LYP distillation, QM40 OODtotal-energy MAE: decreases by 66% on averageImproved
add DI-LossEG-XC · B3LYP distillation, QM40 OODtotal energy error, L2 density: total energy error cut by 80%; L2[rho] lowered by 8%Improved
add DI-Losssemi-local ML-XC functionals · B3LYP distillation, QM40 OODE_rho, dipole and L2 density errors: E_rho 1.2 -> 0.8 mEh; dipole and L2 not uniformly improvedMixed
add gradient term only (without Hessian) to E+rho (ablation)ML-XC functionals · TDDFT excitation MAEmean excitation MAE: raises MAE by 9--14% for XCdiffHurt
add Hessian term on top of gradient term (ablation)XCdiff · TDDFT excitation MAEmean excitation MAE: reduced by 28%, 24%, 24%, 19% for 7, 9, 12, 14 heavy atoms; 35% for 5Improved
add DI-LossXCdiff · TDDFT fifth excitation, 7 heavy atomsMAE (meV): 356 -> 233 meVImproved
add DI-LossXCdiff · TDDFT first excitation, 14 heavy atomsMAE (meV): 462 -> 386 meVImproved
initialize B3LYP with distilled ML-XC density instead of SCAN densityB3LYP SCF · QM9 9 heavy atomsrelative iteration count: about 65% -> about 50% of MINAO iterationsImproved
add derivative supervision to distillation for SCF initialization (ablation)distilled ML-XC · QM9 B3LYP RICrelative iteration countNo clear effect
add Hessian term (ablation)distilled functionals · SCF iterations vs B3LYP referenceSCF iterations: distilled need ~15% more iterations; Hessian recovers partImproved
initialize B3LYP with six NNmGGA cyclesB3LYP on 35-heavy-atom QM40 molecule · QM40walltime speedup: 1.35xImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy