pith. sign in

arxiv: 2606.07239 · v1 · pith:23PLNXS7new · submitted 2026-06-05 · 💻 cs.LG

Generative Molecular Morphing for Flexible-Size Design via Unbalanced Optimal Transport

Malte Franke , Stefan P. Schmid , Zarko Ivkovic , Kjell Jorner , Andreas Krause This is my paper

Pith reviewed 2026-06-27 22:51 UTC · model grok-4.3

classification 💻 cs.LG
keywords generative molecular designunbalanced optimal transportgeometric graphs3D moleculesflexible molecular sizeproperty steeringscaffoldsout-of-distribution generation
0
0 comments X

The pith

Morph generates 3D molecules of variable atom count by applying unbalanced optimal transport to geometric graphs.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents Morph, a generative model that produces 3D molecular structures whose size can change during sampling rather than remaining fixed. It achieves this by using unbalanced optimal transport on geometric graphs, which lets the model incorporate structural priors such as scaffolds and steer generation toward target properties. A sympathetic reader would care because many molecular properties are tied to atom count, so a model that jointly handles size and properties can reach higher-reward designs that fixed-size diffusion and flow models cannot access. Morph is shown to match the performance of current fixed-size state-of-the-art methods while succeeding at out-of-distribution generation.

Core claim

Morph is a flexible-size generative model for conditional and unconditional 3D molecular design based on geometric graphs. By dynamically adapting the number of atoms through unbalanced optimal transport, it integrates existing structural priors and enhances property steering without post-hoc adjustments, matching fixed-size state-of-the-art performance while enabling sampling in regimes where prior models fail.

What carries the argument

Unbalanced optimal transport on geometric graphs, which performs dynamic size adaptation while preserving structural information.

If this is right

  • Molecular generation can now target properties whose optimum depends on atom count without separate size-selection steps.
  • Scaffolds and other partial structures can be used directly as conditioning input rather than requiring post-processing.
  • Steering toward high-reward molecules becomes stronger because size and property distributions are modeled jointly.
  • Generation succeeds in property-size regimes outside the training distribution where fixed-size models produce invalid outputs.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same transport mechanism could be tested on other variable-length structured objects such as proteins or materials lattices.
  • Pairing Morph with reinforcement learning loops might further improve property optimization by exploiting the continuous size flexibility.
  • Efficiency at very large atom counts remains untested and could reveal whether the transport computation scales without additional approximations.

Load-bearing premise

Unbalanced optimal transport on geometric graphs can adapt molecular size dynamically without lowering generation quality or losing the ability to use structural priors.

What would settle it

A side-by-side evaluation on standard fixed-size molecular benchmarks in which Morph produces lower validity, novelty, or property scores than current diffusion or flow models.

Figures

Figures reproduced from arXiv: 2606.07239 by Andreas Krause, Kjell Jorner, Malte Franke, Stefan P. Schmid, Zarko Ivkovic.

Figure 1
Figure 1. Figure 1: Geometric graph generation as a jump-flow process. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Interpolation between a randomly sampled graph ( [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Conditioning on out-of-distribution sizes discovers completely new designs – hinting at [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Conditional design with Morph using 10k generated samples. We condition on different values of logP (dashed red line). The training set density is shown in gray with contour, generated molecules in purple. Although n0 is sampled uniformly, the observed conditional distributions p(n1|logP) closely match the corresponding section of the training distribution. can steer size depending on the property conditio… view at source ↗
Figure 5
Figure 5. Figure 5: Exemplary generation trajectory for the scaffold decoration task. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Multi-property conditioning of Morph on QED and logP. For each combination, we sample 10k molecules and visualize their properties in purple. The training distribution is shown in gray [PITH_FULL_IMAGE:figures/full_fig_p020_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Randomly selected QM9 interpolation trajectories. [PITH_FULL_IMAGE:figures/full_fig_p021_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Sampled valid molecules using Morph trained on QM9. 22 [PITH_FULL_IMAGE:figures/full_fig_p022_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Sampled valid molecules using Morph trained on GEOM-Drugs. 23 [PITH_FULL_IMAGE:figures/full_fig_p023_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Randomly sampled valid molecules using ncond = 30 [PITH_FULL_IMAGE:figures/full_fig_p024_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Randomly sampled invalid molecules using ncond = 30 [PITH_FULL_IMAGE:figures/full_fig_p024_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Randomly sampled valid molecules using ncond = 35 24 [PITH_FULL_IMAGE:figures/full_fig_p024_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Randomly sampled invalid molecules using ncond = 35 [PITH_FULL_IMAGE:figures/full_fig_p025_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Randomly sampled valid molecules using ncond = 40 [PITH_FULL_IMAGE:figures/full_fig_p025_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Randomly sampled invalid molecules using ncond = 40 25 [PITH_FULL_IMAGE:figures/full_fig_p025_15.png] view at source ↗
read the original abstract

The success of generative molecular design hinges on a model's steerability toward high-reward samples. Because many molecular properties are intrinsically linked to molecular size, accurately capturing the joint distribution of properties and the number of atoms is essential. However, current diffusion and flow-based models fix the number of atoms, which ultimately limits their ability to navigate this complex relationship. To address this, we introduce Morph, a flexible-size generative model for conditional and unconditional 3D molecular design based on geometric graphs. By dynamically adapting size, Morph can seamlessly integrate existing structural priors, like scaffolds, and significantly enhances property steering. We show that Morph matches current fixed-size state-of-the-art models while offering the benefit of unparalleled sampling flexibility. We demonstrate out-of-distribution generation in regimes where previous models fail, paving the way for enhanced generative modeling for molecular design.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 2 minor

Summary. The paper introduces Morph, a generative model for conditional and unconditional 3D molecular design that operates on geometric graphs via unbalanced optimal transport. Unlike prior diffusion and flow-based approaches that fix the number of atoms, Morph dynamically adapts molecular size. The central claims are that it matches the generation quality of fixed-size state-of-the-art models, seamlessly incorporates structural priors such as scaffolds, improves property steering, and enables out-of-distribution generation in regimes where previous models fail.

Significance. If the empirical claims are substantiated, the work would represent a meaningful advance in generative molecular design. Allowing variable atom counts directly addresses the coupling between molecular size and many target properties, potentially improving steerability and practical applicability without post-hoc fixes. The use of unbalanced OT on geometric graphs to achieve this flexibility while preserving structural priors is a technically interesting direction.

minor comments (2)
  1. [Abstract / Results] The abstract states that Morph 'matches current fixed-size state-of-the-art models' on generation quality; the results section should include direct quantitative comparisons (e.g., validity, uniqueness, property optimization metrics) against the specific baselines referenced.
  2. [Method] Clarify how the unbalanced OT formulation is implemented to enforce size adaptation without introducing additional hyperparameters that could affect the 'parameter-free' aspects of the structural priors.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive evaluation of the manuscript and for recommending minor revision. We are encouraged by the recognition that addressing variable molecular size via unbalanced optimal transport on geometric graphs represents a meaningful technical direction with potential practical benefits for steerability and out-of-distribution generation.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper introduces Morph as a new generative model based on unbalanced optimal transport applied to geometric graphs, claiming it enables flexible-size sampling while matching fixed-size SOTA performance and enabling OOD generation. The abstract and provided claims present this as an empirical method with stated benefits, without any equations, derivations, or performance assertions that reduce by construction to fitted inputs, self-definitions, or self-citation chains. No load-bearing step is shown to rename a known result, smuggle an ansatz, or import uniqueness from prior author work as an external theorem. The central claims rest on the described OT formulation and experimental demonstrations rather than tautological reductions, making the derivation self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no explicit free parameters, axioms, or invented entities; the central claim rests on the unverified assertion that unbalanced OT solves the size-property joint distribution problem.

pith-pipeline@v0.9.1-grok · 5682 in / 1031 out tokens · 19119 ms · 2026-06-27T22:51:14.570927+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

58 extracted references · 5 canonical work pages

  1. [1]

    Machine learning-aided generative molecular design.Nature Machine Intelligence, 6(6):589–604, 2024

    Yuanqi Du, Arian R Jamasb, Jeff Guo, Tianfan Fu, Charles Harris, Yingheng Wang, Chenru Duan, Pietro Liò, Philippe Schwaller, and Tom L Blundell. Machine learning-aided generative molecular design.Nature Machine Intelligence, 6(6):589–604, 2024

    2024

  2. [2]

    Midi: Mixed graph and 3d denoising diffusion for molecule generation

    Clement Vignac, Nagham Osman, Laura Toni, and Pascal Frossard. Midi: Mixed graph and 3d denoising diffusion for molecule generation. InJoint European Conference on Machine Learning and Knowledge Discovery in Databases, pages 560–576. Springer, 2023

    2023

  3. [3]

    Navigating the design space of equivariant diffusion-based generative models for de novo 3d molecule generation.arXiv preprint arXiv:2309.17296, 2023

    Tuan Le, Julian Cremer, Frank Noe, Djork-Arné Clevert, and Kristof Schütt. Navigating the design space of equivariant diffusion-based generative models for de novo 3d molecule generation.arXiv preprint arXiv:2309.17296, 2023

    arXiv 2023

  4. [4]

    Flowmol3: flow matching for 3d de novo small-molecule generation.Digital Discovery, 5(5):2052–2066, 2026

    Ian Dunn and David R Koes. Flowmol3: flow matching for 3d de novo small-molecule generation.Digital Discovery, 5(5):2052–2066, 2026

    2052

  5. [5]

    SemlaFlow – efficient 3d molecular generation with latent attention and equivariant flow matching.arXiv preprint arXiv:2406.07266, 2025

    Ross Irwin, Alessandro Tibo, Jon Paul Janet, and Simon Olsson. SemlaFlow – efficient 3d molecular generation with latent attention and equivariant flow matching.arXiv preprint arXiv:2406.07266, 2025

    arXiv 2025

  6. [6]

    Marton Havasi, Brian Karrer, Itai Gat, and Ricky T. Q. Chen. Edit flows: Flow matching with edit operations.arXiv preprint arXiv:2506.09018, 2025

    arXiv 2025

  7. [7]

    Peter Holderrieth, Marton Havasi, Jason Yim, Neta Shaul, Itai Gat, Tommi Jaakkola, Brian Karrer, Ricky T. Q. Chen, and Yaron Lipman. Generator matching: Generative modeling with arbitrary markov processes.arXiv preprint arXiv:2410.20587, 2025

    arXiv 2025

  8. [8]

    Cambridge university press, 2009

    David Applebaum.Lévy processes and stochastic calculus. Cambridge university press, 2009

    2009

  9. [9]

    John Wiley & Sons, 2009

    Stewart N Ethier and Thomas G Kurtz.Markov processes: characterization and convergence. John Wiley & Sons, 2009

    2009

  10. [10]

    Approximate graph edit distance computation by means of bipartite graph matching.Image and Vision computing, 27(7):950–959, 2009

    Kaspar Riesen and Horst Bunke. Approximate graph edit distance computation by means of bipartite graph matching.Image and Vision computing, 27(7):950–959, 2009

    2009

  11. [11]

    Unbalanced optimal transport, from theory to numerics.Handbook of Numerical Analysis, 24:407–471, 2023

    Thibault Séjourné, Gabriel Peyré, and François-Xavier Vialard. Unbalanced optimal transport, from theory to numerics.Handbook of Numerical Analysis, 24:407–471, 2023

    2023

  12. [12]

    The hungarian method for the assignment problem.Naval research logistics quarterly, 2(1-2):83–97, 1955

    Harold W Kuhn. The hungarian method for the assignment problem.Naval research logistics quarterly, 2(1-2):83–97, 1955

    1955

  13. [13]

    Algorithms for the assignment and transportation problems.Journal of the society for industrial and applied mathematics, 5(1):32–38, 1957

    James Munkres. Algorithms for the assignment and transportation problems.Journal of the society for industrial and applied mathematics, 5(1):32–38, 1957. 10

    1957

  14. [14]

    A solution for the best rotation to relate two sets of vectors.Foundations of Crystallography, 32(5):922–923, 1976

    Wolfgang Kabsch. A solution for the best rotation to relate two sets of vectors.Foundations of Crystallography, 32(5):922–923, 1976

    1976

  15. [15]

    Equivariant flow matching.Advances in Neural Information Processing Systems, 36:59886–59910, 2023

    Leon Klein, Andreas Krämer, and Frank Noé. Equivariant flow matching.Advances in Neural Information Processing Systems, 36:59886–59910, 2023

    2023

  16. [16]

    Itai Gat, Tal Remez, Neta Shaul, Felix Kreuk, Ricky T. Q. Chen, Gabriel Synnaeve, Yossi Adi, and Yaron Lipman. Discrete flow matching.arXiv preprint arXiv:2407.15595, 2024

    arXiv 2024

  17. [17]

    Gener- ative flows on discrete state-spaces: Enabling multimodal flows with applications to protein co-design.arXiv preprint arXiv:2402.04997, 2024

    Andrew Campbell, Jason Yim, Regina Barzilay, Tom Rainforth, and Tommi Jaakkola. Gener- ative flows on discrete state-spaces: Enabling multimodal flows with applications to protein co-design.arXiv preprint arXiv:2402.04997, 2024

    arXiv 2024

  18. [18]

    Flow matching for generative modeling.arXiv preprint arXiv:2210.02747, 2022

    Yaron Lipman, Ricky TQ Chen, Heli Ben-Hamu, Maximilian Nickel, and Matt Le. Flow matching for generative modeling.arXiv preprint arXiv:2210.02747, 2022

    Pith/arXiv arXiv 2022

  19. [19]

    John Nguyen, Marton Havasi, Tariq Berrada, Luke Zettlemoyer, and Ricky T. Q. Chen. One- Flow: Concurrent mixed-modal and interleaved generation with edit flows.arXiv preprint arXiv:2510.03506, 2025

    arXiv 2025

  20. [20]

    Enumeration of 166 billion organic small molecules in the chemical universe database gdb-17.Journal of chemical information and modeling, 52(11):2864–2875, 2012

    Lars Ruddigkeit, Ruud Van Deursen, Lorenz C Blum, and Jean-Louis Reymond. Enumeration of 166 billion organic small molecules in the chemical universe database gdb-17.Journal of chemical information and modeling, 52(11):2864–2875, 2012

    2012

  21. [21]

    Quantum chemistry structures and properties of 134 kilo molecules.Scientific Data, 1, 2014

    Raghunathan Ramakrishnan, Pavlo O Dral, Matthias Rupp, and O Anatole von Lilienfeld. Quantum chemistry structures and properties of 134 kilo molecules.Scientific Data, 1, 2014

    2014

  22. [22]

    Geom, energy-annotated molecular conforma- tions for property prediction and molecular generation.Scientific data, 9(1):185, 2022

    Simon Axelrod and Rafael Gomez-Bombarelli. Geom, energy-annotated molecular conforma- tions for property prediction and molecular generation.Scientific data, 9(1):185, 2022

    2022

  23. [23]

    Morris, and Charlotte M

    Martin Buttenschoen, Garrett M. Morris, and Charlotte M. Deane. PoseBusters: AI-based docking methods fail to generate physically valid poses or generalise to novel sequences. Chemical Science, 15:3130–3139, 2024. doi: 10.1039/D3SC04185A. URL http://dx.doi. org/10.1039/D3SC04185A

    work page doi:10.1039/d3sc04185a 2024

  24. [24]

    Tadmor, Richard G

    Cheng Zeng, Jirui Jin, Connor Ambrose, George Karypis, Mark Transtrum, Ellad B. Tadmor, Richard G. Hennig, Adrian Roitberg, Stefano Martiniani, and Mingjie Liu. PropMolFlow: property-guided molecule generation with geometry-complete flow matching.Nature Computa- tional Science, pages 1–10, January 2026. ISSN 2662-8457. doi: 10.1038/s43588-025-00946-y. URL...

    work page doi:10.1038/s43588-025-00946-y 2026

  25. [25]

    Rdkit: Open-source cheminformatics, 2026

    Greg Landrum et al. Rdkit: Open-source cheminformatics, 2026

    2026

  26. [26]

    Libinvent: Reaction-based generative scaffold decoration for in silico library design.Journal of Chemical Information and Modeling, 62(9):2046– 2063, 2022

    Vendy Fialková, Jiaxi Zhao, Kostas Papadopoulos, Ola Engkvist, Esben Jannik Bjerrum, Thierry Kogej, and Atanas Patronov. Libinvent: Reaction-based generative scaffold decoration for in silico library design.Journal of Chemical Information and Modeling, 62(9):2046– 2063, 2022. doi: 10.1021/acs.jcim.1c00469. URL https://doi.org/10.1021/acs.jcim. 1c00469. PM...

    work page doi:10.1021/acs.jcim.1c00469 2046

  27. [27]

    Structure-based drug design with equivariant diffusion models.Nature Computational Science, 4(12):899–909, 2024

    Arne Schneuing, Charles Harris, Yuanqi Du, Kieran Didi, Arian Jamasb, Ilia Igashov, Weitao Du, Carla Gomes, Tom L Blundell, Pietro Lio, Max Welling, Michael Bronstein, and Bruno Correia. Structure-based drug design with equivariant diffusion models.Nature Computational Science, 4(12):899–909, 2024

    2024

  28. [28]

    Levenshtein transformer.arXiv preprint arXiv:1905.11006, 2019

    Jiatao Gu, Changhan Wang, and Jake Zhao. Levenshtein transformer.arXiv preprint arXiv:1905.11006, 2019

    arXiv 1905

  29. [29]

    Insertion-deletion transformer

    Laura Ruis, Mitchell Stern, Julia Proskurnia, and William Chan. Insertion-deletion transformer. arXiv preprint arXiv:2001.05540, 2020

    arXiv 2001

  30. [30]

    Johnson, Jacob Austin, Rianne van den Berg, and Daniel Tarlow

    Daniel D. Johnson, Jacob Austin, Rianne van den Berg, and Daniel Tarlow. Beyond in- place corruption: Insertion and deletion in denoising probabilistic models.arXiv preprint arXiv:2107.07675, 2021. 11

    arXiv 2021

  31. [31]

    Any-order flexible length masked diffusion

    Jaeyeon Kim, Lee Cheuk-Kit, Carles Domingo-Enrich, Yilun Du, Sham Kakade, Timothy Ngotiaoco, Sitan Chen, and Michael Albergo. Any-order flexible length masked diffusion. arXiv preprint arXiv:2509.01025, 2025

    arXiv 2025

  32. [32]

    Insertion language models: Sequence generation with arbitrary-position insertions.arXiv preprint arXiv:2505.05755, 2025

    Dhruvesh Patel, Aishwarya Sahoo, Avinash Amballa, Tahira Naseem, Tim GJ Rudner, and Andrew McCallum. Insertion language models: Sequence generation with arbitrary-position insertions.arXiv preprint arXiv:2505.05755, 2025

    arXiv 2025

  33. [33]

    Graphrnn: Generat- ing realistic graphs with deep auto-regressive models

    Jiaxuan You, Rex Ying, Xiang Ren, William Hamilton, and Jure Leskovec. Graphrnn: Generat- ing realistic graphs with deep auto-regressive models. InInternational conference on machine learning, pages 5708–5717. PMLR, 2018

    2018

  34. [34]

    Efficient graph generation with graph recurrent attention networks

    Renjie Liao, Yujia Li, Yang Song, Shenlong Wang, Will Hamilton, David K Duvenaud, Raquel Urtasun, and Richard Zemel. Efficient graph generation with graph recurrent attention networks. Advances in neural information processing systems, 32, 2019

    2019

  35. [35]

    Digress: Discrete denoising diffusion for graph generation.arXiv preprint arXiv:2209.14734, 2022

    Clement Vignac, Igor Krawczuk, Antoine Siraudin, Bohan Wang, V olkan Cevher, and Pas- cal Frossard. Digress: Discrete denoising diffusion for graph generation.arXiv preprint arXiv:2209.14734, 2022

    arXiv 2022

  36. [36]

    Score-based generative modeling of graphs via the system of stochastic differential equations

    Jaehyeong Jo, Seul Lee, and Sung Ju Hwang. Score-based generative modeling of graphs via the system of stochastic differential equations. InInternational conference on machine learning, pages 10362–10383. PMLR, 2022

    2022

  37. [37]

    Graph diffusion that can insert and delete

    Matteo Ninniri, Marco Podda, and Davide Bacciu. Graph diffusion that can insert and delete. arXiv preprint arXiv:2506.15725, 2025

    arXiv 2025

  38. [38]

    Equivariant diffusion for molecule generation in 3d.International conference on machine learning, pages 8867–8887, 2022

    Emiel Hoogeboom, Vıctor Garcia Satorras, Clément Vignac, and Max Welling. Equivariant diffusion for molecule generation in 3d.International conference on machine learning, pages 8867–8887, 2022

    2022

  39. [39]

    Joshi, Xiang Fu, Yi-Lun Liao, Vahe Gharakhanyan, Benjamin Kurt Miller, Anuroop Sriram, and Zachary W

    Chaitanya K. Joshi, Xiang Fu, Yi-Lun Liao, Vahe Gharakhanyan, Benjamin Kurt Miller, Anuroop Sriram, and Zachary W. Ulissi. All-atom diffusion transformers: Unified generative modelling of molecules and materials.arXiv preprint arXiv:2503.03965, 2025

    arXiv 2025

  40. [40]

    Benjamin Erichson, and Michael W

    Alex Morehead, Miruna Cretu, Antonia Panescu, Rishabh Anand, Maurice Weiler, Tynan Perez, Samuel Blau, Steven Farrell, Wahid Bhimji, Anubhav Jain, Hrushikesh Sahasrabuddhe, Pietro Lio, Tommi Jaakkola, Rafael Gomez-Bombarelli, Rex Ying, N. Benjamin Erichson, and Michael W. Mahoney. Zatom-1: A multimodal flow foundation model for 3d molecules and materials....

    Pith/arXiv arXiv 2026

  41. [41]

    Diffdec: Structure-aware scaffold decoration with an end-to-end diffusion model.Journal of Chemical Information and Modeling, 64(7):2554–2564, 2024

    Junjie Xie, Sheng Chen, Jinping Lei, and Yuedong Yang. Diffdec: Structure-aware scaffold decoration with an end-to-end diffusion model.Journal of Chemical Information and Modeling, 64(7):2554–2564, 2024. doi: 10.1021/acs.jcim.3c01466. URL https://doi.org/10.1021/ acs.jcim.3c01466. PMID: 38267393

    work page doi:10.1021/acs.jcim.3c01466 2024

  42. [42]

    TABASCO: A fast, sim- plified model for molecular generation with improved physical quality.arXiv preprint arXiv:2507.00899, 2025

    Carlos V onessen, Charles Harris, Miruna Cretu, and Pietro Liò. TABASCO: A fast, sim- plified model for molecular generation with improved physical quality.arXiv preprint arXiv:2507.00899, 2025

    arXiv 2025

  43. [43]

    Applications of modular co-design for de novo 3d molecule generation.Digital Discovery, 5(2):754–768, 2026

    Danny Reidenbach, Filipp Nikitin, Olexandr Isayev, and Saee Gopal Paliwal. Applications of modular co-design for de novo 3d molecule generation.Digital Discovery, 5(2):754–768, 2026

    2026

  44. [44]

    Equivariant 3d-conditional diffusion model for molecular linker design.Nature Machine Intelligence, 6(4):417–427, 2024

    Ilia Igashov, Hannes Stärk, Clément Vignac, Arne Schneuing, Victor Garcia Satorras, Pascal Frossard, Max Welling, Michael Bronstein, and Bruno Correia. Equivariant 3d-conditional diffusion model for molecular linker design.Nature Machine Intelligence, 6(4):417–427, 2024

    2024

  45. [45]

    Prior-guided flow matching for target-aware molecule design with learnable atom number.arXiv preprint arXiv: 2509.01486, 2025

    Jingyuan Zhou, Hao Qian, Shikui Tu, and Lei Xu. Prior-guided flow matching for target-aware molecule design with learnable atom number.arXiv preprint arXiv: 2509.01486, 2025

    arXiv 2025

  46. [46]

    Trans-dimensional generative modeling via jump diffusion models.arXiv preprint arXiv:2305.16261, 2023

    Andrew Campbell, William Harvey, Christian Weilbach, Valentin De Bortoli, Tom Rainforth, and Arnaud Doucet. Trans-dimensional generative modeling via jump diffusion models.arXiv preprint arXiv:2305.16261, 2023. 12

    arXiv 2023

  47. [47]

    Cheng, Chong Sun, and Alán Aspuru-Guzik

    Austin H. Cheng, Chong Sun, and Alán Aspuru-Guzik. Scalable autoregressive 3d molecule generation.arXiv preprint arXiv:2505.13791, 2025

    arXiv 2025

  48. [48]

    Joshi, Niklas W

    Rajendra P. Joshi, Niklas W. A. Gebauer, Mridula Bontha, Mercedeh Khazaieli, Rhema M. James, James B. Brown, and Neeraj Kumar. 3d-scaffold: A deep learning framework to generate 3d coordinates of drug-like molecules with desired scaffolds.The Journal of Physical Chemistry B, 125(44):12166–12176, 2021. doi: 10.1021/acs.jpcb.1c06437. URL https: //doi.org/10...

    work page doi:10.1021/acs.jpcb.1c06437 2021

  49. [49]

    Symphony: Symmetry- equivariant point-centered spherical harmonics for 3d molecule generation

    Ameya Daigavane, Song Eun Kim, Mario Geiger, and Tess Smidt. Symphony: Symmetry- equivariant point-centered spherical harmonics for 3d molecule generation. InInternational Conference on Learning Representations, volume 2024, pages 33975–34002, 2024

    2024

  50. [50]

    Neat: Neighborhood- guided, efficient, autoregressive set transformer for 3d molecular generation.arXiv preprint arXiv:2512.05844, 2025

    Daniel Rose, Roxane Axel Jacob, Johannes Kirchmair, and Thierry Langer. Neat: Neighborhood- guided, efficient, autoregressive set transformer for 3d molecular generation.arXiv preprint arXiv:2512.05844, 2025

    Pith/arXiv arXiv 2025

  51. [51]

    Branching flows: Discrete, continuous, and manifold flow matching with splits and deletions.arXiv preprint arXiv:2511.09465, 2025

    Lukas Billera, Hedwig Nora Nordlinder, Jack Collier Ryder, Anton Oresten, Aron Stålmarck, Theodor Mosetti Björk, and Ben Murrell. Branching flows: Discrete, continuous, and manifold flow matching with splits and deletions.arXiv preprint arXiv:2511.09465, 2025

    Pith/arXiv arXiv 2025

  52. [52]

    3d molecule generation from rigid motifs via SE(3) flows.arXiv preprint arXiv:2601.16955, 2026

    Roman Poletukhin, Marcel Kollovieh, Eike Eberhard, and Stephan Günnemann. 3d molecule generation from rigid motifs via SE(3) flows.arXiv preprint arXiv:2601.16955, 2026

    arXiv 2026

  53. [53]

    Pinheiro, Arian Jamasb, Omar Mahmood, Vishnu Sresht, and Saeed Saremi

    Pedro O. Pinheiro, Arian Jamasb, Omar Mahmood, Vishnu Sresht, and Saeed Saremi. Structure- based drug design by denoising voxel grids.arXiv preprint arXiv:2405.03961, 2024

    arXiv 2024

  54. [54]

    Pinheiro, Joshua Rackers, Joseph Kleinhenz, Michael Maser, Omar Mahmood, An- drew Martin Watkins, Stephen Ra, Vishnu Sresht, and Saeed Saremi

    Pedro O. Pinheiro, Joshua Rackers, Joseph Kleinhenz, Michael Maser, Omar Mahmood, An- drew Martin Watkins, Stephen Ra, Vishnu Sresht, and Saeed Saremi. 3d molecule generation by denoising voxel grids.arXiv preprint arXiv:2306.07473, 2024

    arXiv 2024

  55. [55]

    ProxelGen: Generating proteins as 3d densities.arXiv preprint arXiv:2506.19820, 2025

    Felix Faltings, Hannes Stark, Regina Barzilay, and Tommi Jaakkola. ProxelGen: Generating proteins as 3d densities.arXiv preprint arXiv:2506.19820, 2025

    arXiv 2025

  56. [56]

    CORDS: Continuous representations of discrete structures.arXiv preprint arXiv:2601.21583, 2026

    Tin Hadži Veljkovi´c, Erik Bekkers, Michael Tiemann, and Jan-Willem van de Meent. CORDS: Continuous representations of discrete structures.arXiv preprint arXiv:2601.21583, 2026. A Technical Appendices and Supplementary Material A.1 Data details QM9The QM9 dataset [ 21] contains around 134k small organic molecules with up to 9 heavy atoms of types { H, C, ...

    arXiv 2026

  57. [57]

    Invariant Scalar Predictions For each node i, the invariant features hi ∈R d are mapped to the scalar Gaussian mixture parameters via a Multi-Layer Perceptron (MLP): [πi, σi, ai, ci] =Split(MLP scalar(hi))(26) Appropriate activations are applied to the partitions: Softmax to obtain the K mixture weights πi, atom type probabilities Ai, and charge type prob...

  58. [58]

    drug-likeness

    Equivariant Mean Predictions To compute the equivariant GMM means µi,k ∈R 3, we predict spatial weights wij ∈R K for each neighbor j∈ N(i) . This is done using the source and target invariant features, along with a Radial Basis Function (RBF) embedding of the Euclidean distanced ij =∥x i −x j∥2: wij =γtanh MLPcoord hi ∥h j ∥RBF(d ij) (27) where ∥ denotes ...

    2000