AI Paper Atlas

Research map / Probabilistic and scientific modeling / Molecules and proteins

EvoEGF-Mol: Evolving Exponential Geodesic Flow for Structure-based Drug Design

Yaowei Jin, Junjie Wang, Cheng Cao, Penglei Wang, Duo An, Qian Shi

ICML 2026 · Probabilistic and scientific modeling › Molecules and proteins · approach: Structure-based 3D molecule generation · 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
use evolving endpoint instead of fixed Dirac target e-geodesic (ablation)EvoEGF-Mol · CrossDock test (100 proteins)PB-Valid, strain energy, VinaImproved
use designed Dirichlet evolution instead of standard Dirichlet flow matching concentration (ablation)EvoEGF-Mol · CrossDock test (100 proteins)strain energy, PB-ValidImproved
use proposed Dirichlet parameterization instead of uniform-to-one-hot schedule (ablation)EvoEGF-Mol · CrossDock test (100 proteins)performanceImproved
set smoothing coefficient lambda to 0.2 versus smaller or larger (ablation)EvoEGF-Mol · CrossDock test (100 proteins)geometric quality and stabilityImproved
add explicit bond diffusion during training (ablation)EvoEGF-Mol · CrossDock test (100 proteins)PB-Valid, strain energyImproved
reduce sampling steps from 100 to 50 (ablation)EvoEGF-Mol · CrossDock test (100 proteins)PB-Valid: 92.0%No clear effect
use EvoEGF-Mol instead of MolCRAFTEvoEGF-Mol · CrossDock test (100 proteins)PB-Valid (%): 84.6 -> 93.4Improved
use EvoEGF-Mol instead of MolCRAFTEvoEGF-Mol · MolGenBench (CrossDock proteins)scaffold pass rate (%): 23.07 -> 31.79Improved
use EvoEGF-Mol instead of MolCRAFTEvoEGF-Mol · CrossDock test (100 proteins)Vina Score avg: -6.55 -> -6.14Hurt
use EvoEGF-Mol instead of DecompDiffEvoEGF-Mol · MolGenBench (CrossDock proteins)Hit Rate (%): 4.99 -> 3.88Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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