Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Tokenizing 3D Molecule Structure with Quantized Spherical Coordinates

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Treating each atom's local position as a discrete spherical-coordinate token turns 3D molecule generation into a language-model task and runs about 28 times faster than diffusion.

desk verdict Mol-StrucTok is a fresh, practical 3D tokenization method with a real speed advantage, but the unhandled collinearity degeneracy and missing error bars require revision before I'd trust the headline numbers. read the letter →

arxiv 2412.01564 v1 pith:NHBJILOL submitted 2024-12-02 cs.LG q-bio.BM

Kaiyuan Gao , Yusong Wang , Haoxiang Guan , Zun Wang , Qizhi Pei , John E. Hopcroft , Kun He , Lijun Wu This is my paper · ORCID
classification cs.LGq-bio.BM
keywords 3DmoleculegenerationtokenizationsphericalcoordinatesvectorquantizationlanguagemodelsSE(3)invarianceQM9propertypredictionSELFIES
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper sets out to make 3D molecular structure as natural for language models as SMILES strings are for 2D molecules, and claims this can be done by quantizing each atom's position into discrete tokens rather than predicting continuous coordinates. It introduces a spherical line notation: every atom token carries a local distance $d_i$, polar angle $\theta_i$, and azimuth $\varphi_i$ measured in a frame built from a focal atom and two reference atoms, which makes the coordinates invariant under rotation and translation. A VQ-VAE turns these continuous coordinates, together with local bond lengths and angles, into one of 256 codebook tokens, and a GPT-2 style model is trained to generate the resulting token sequence. The paper reports that the model produces valid, chemically stable molecules, matches diffusion methods in quality, generates them about 28 times faster, and conditions more accurately on quantum properties.

What carries the argument

The load-bearing object is the per-atom spherical descriptor $z_i = (d_i, \theta_i, |\varphi_i|, \mathrm{sign}(\varphi_i), l_{j_1}, l_{j_2}, l_{j_3}, l_{j_4}, \alpha_{j_1 i j_2}, \ldots)$, a 14-dimensional vector that is invariant under rotations and translations by construction. It carries the argument in two ways. First, it linearizes a 3D structure: when appended to the atom tokens of SMILES or SELFIES, the sequence of descriptors determines the full geometry. Second, the VQ-VAE maps it to one of 256 codebook entries, giving the discrete structural alphabet that a GPT-2 model can predict autoregressively. The reference-frame construction uses topology-based selection, where the focal atom is the closest already-written neighbor in the molecular graph and its references are that neighbor's predecessors, which the paper shows is more robust to coordinate noise than 1D sequence-based or 3D distance-based selection.

What would settle it

Run the tokenizer's descriptor extraction on a linear triatomic molecule such as CO₂: at atoms where the focal atom and its two reference atoms are collinear, the Gram-Schmidt denominator in the construction of $e_2$ becomes exactly zero, so no valid $(d_i, \theta_i, \varphi_i)$ descriptor exists and the pipeline either crashes, assigns an arbitrary frame, or fails to reconstruct the molecule.

Watch

Extended reading notes

Core claim

The central discovery is that an SE(3)-invariant discretization of local geometry is sufficient to reduce 3D molecule generation to next-token prediction without sacrificing chemical validity. For each atom, Mol-StrucTok appends to its ordinary line-notation token a 14-dimensional descriptor: the distance $d_i$, polar angle $\theta_i$, azimuth $\varphi_i$, and its sign, measured in a local frame fixed by a focal atom and two reference atoms chosen by 2D topology, followed by the bond lengths and bond angles to the atom's four nearest neighbors. A VQ-VAE with a 256-entry codebook quantizes that descriptor into a single discrete token, so the vocabulary is the product of atom types and structural alphabet entries. Trained as a GPT-2 next-token predictor on QM9, the paper claims 98.02% validity by lookup table, 88.30% molecular stability, and a 39.8 samples per second generation throughput versus 1.4 for the diffusion baselines, a roughly 28 times speedup; in conditional generation it reports a gap-energy MAE of 89 meV, where the prior best baseline GeoBFN reports 577 meV.

Load-bearing premise

The construction assumes that for every atom, its focal atom and two reference atoms are not collinear; when they are, the local spherical coordinates are undefined and the whole tokenization has nothing to attach to the atom token.

Editorial extensions

If this is right

  • Molecule generation can ride on the same infrastructure as large language models, including hardware-accelerated autoregressive decoding and KV-cache, which is why the reported throughput reaches about 40 samples per second on one A100 GPU.
  • Property-conditioned generation becomes a conditional language-model problem; on QM9 the method reports a gap-energy MAE of 89 meV against 577 meV for GeoBFN, suggesting tighter control of target properties.
  • The discrete structure tokens embed into a 2D graph transformer and improve QM9 property prediction for HOMO, LUMO, and their gap, so the tokenizer is useful beyond generation.
  • The representation is not tied to one line notation; switching from SMILES to SELFIES keeps validity and stability nearly unchanged, so any line notation can carry the 3D tokens.
  • A direct practical consequence of the speed advantage is that screening large candidate libraries by generating 3D structures becomes feasible at much lower sampling cost than diffusion-based generation.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper leaves implicit that its discrete interface should accept any text-style conditioning, not just a scalar property; a testable extension is to prepend natural-language or protein-pocket constraints and fine-tune the same GPT-2 model.
  • The 256-token codebook is a likely diversity bottleneck, since uniqueness (about 85%) trails diffusion baselines; enlarging the codebook or adding stochastic decoding is an untested way to close that gap.
  • No degenerate-case handling is given for collinear reference atoms; until a fallback frame is defined, linear molecules are a boundary case the tokenizer likely cannot represent.
  • Because the tokenizer is SE(3)-invariant and line-notation agnostic, it should transfer to conformer generation or to other small-molecule datasets; the paper only demonstrates QM9, so that generality is an inference.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper proposes Mol-StrucTok, a method that converts 3D molecular conformations into discrete tokens by appending local spherical coordinates to an existing 2D line notation (SMILES or SELFIES) and then applying VQ-VAE quantization. The resulting tokens are used as a structure-aware vocabulary for a GPT-2 style autoregressive generator, and as additional embeddings for Graphormer on QM9 property prediction. The authors report competitive validity and stability, a roughly 28x sampling speedup over diffusion baselines, and large improvements in conditional generation MAE over GeoBFN.

Significance. If the construction is made fully well-defined, the paper would be a useful contribution: it provides a way to adapt language-model machinery to 3D molecular structure, with a concrete tokenization scheme and a separation between tokenizer training on PCQM4Mv2 and downstream evaluation on QM9, which avoids obvious circularity. The robustness comparison of three reference-frame choices in Section 5 is a nice empirical sanity check. However, the central tokenization has correctness gaps (degenerate local frames and atoms with fewer than four neighbors), and the headline conditional-generation results need additional scrutiny before the claims can be accepted.

major comments (4)
  1. [Section 3.2 and Appendix A, Eqs. (17)-(18)] The local frame is undefined whenever the reference atoms v_f, v_c1, and v_c2 are collinear, because the Gram-Schmidt numerator x_c2f - (x_c2f·e1)e1 becomes exactly zero and e2 is undefined. With the 2D topology-based rule of Eq. (5), this occurs for linear fragments such as terminal alkynes, allene heavy-atom chains, and CO2. The paper neither mentions this degeneracy nor provides a fallback rule (e.g., skipping to the next non-collinear topological neighbor or using a fixed perpendicular vector). Since the descriptor z_i in Eq. (9) and every downstream token depend on this frame, the tokenization is not well-defined on a nonempty class of valid molecules. The authors must specify a degeneracy-handling strategy and either prove such cases do not occur in their datasets or report how they are processed in the released implementation.
  2. [Section 3.3, Eq. (8)] The understanding descriptor u_i assumes that every atom has at least four nearest neighbors, but molecules in QM9 and PCQM4Mv2 contain terminal atoms (including hydrogen) and linear molecules such as CO2 whose central atom has only two neighbors. The paper does not describe how the missing bond lengths and bond angles are encoded (padding, sentinel values, or exclusion). This is a load-bearing domain restriction for the VQ-VAE input, and the reported reconstruction and generation numbers cannot be interpreted without knowing how these cases are handled.
  3. [Section 4.3, Table 3] The conditional-generation MAEs reported for Mol-StrucTok approach the oracle row (e.g., 0.33 vs 0.10 for alpha, 89 vs 64 meV for Delta-epsilon), while unconditional uniqueness is substantially lower than baselines in Table 1 (85.35% vs 98.85% for GeoLDM). This pattern is consistent with the generator reproducing near-training-set molecules rather than performing genuine condition-controlled sampling. To support the claim of 'significant improvements over GeoBFN', the paper should report variance over seeds, the distance of generated molecules to the training set (e.g., nearest-neighbor in the descriptor space), and the diversity of condition-matched molecules.
  4. [Section 4.4, Table 4] The reported gains from adding Mol-StrucTok embeddings to Graphormer are small and are presented without error bars or significance tests: epsilon_HOMO 46 to 42 meV, epsilon_LUMO 47 to 39 meV, Delta-epsilon 66 to 62 meV. Given that the base Graphormer numbers are only single runs, these differences may be within run-to-run noise. Paired runs with variance estimates are needed before claiming 'consistent improvements'.
minor comments (6)
  1. [Section 3.2, Eq. (5)] The notation N(i,j) is used in the argmax as though it were an indicator of topological adjacency, but it is defined only as 'topological neighbors of atom vi with respect to atom vj'; please replace it with an explicit indicator function or a clear definition.
  2. [Section 3.2, Eq. (6)] The text says the two reference atoms are the closest to the focal atom, but the formula uses argmax over Euclidean distance, which selects the farthest atoms. This is contradictory and should be corrected to argmin if the text is meant literally.
  3. [Section 3.3, Eq. (9)] The descriptor z_i is said to lie in R^14, but it includes sign(phi_i), which is a binary categorical variable; the encoding of this sign (as +1/-1, 0/1, or a one-hot vector) should be stated explicitly.
  4. [Appendix B.2, Eqs. (33) and (38)] The rotation-invariance derivation contains a confusing chain in Eq. (33) where primed and unprimed vectors are mixed; additionally, for phi_i, equality of cosines alone does not determine the angle on (-pi, pi], so the proof should state explicitly that proper rotations preserve the orientation of the frame and that the sign component of the descriptor is handled consistently.
  5. [Appendix C.2] The sentence 'We applied a repetition penalty of 1' is ambiguous, since a repetition penalty of 1 typically means no penalty; please clarify the value and its effect.
  6. [Table 2] The method name is typeset inconsistently as 'MOL-STRUCTOK' in Table 2 while the rest of the paper uses 'Mol-StrucTok'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Mol-StrucTok's tokenizer, generator, and evaluator are not defined in terms of the quantities they predict.

full rationale

The core derivation is self-contained. The structural codebook is trained on PCQM4Mv2 and then frozen; the QM9 generation and property-prediction experiments use this fixed tokenizer, so the codebook is not fitted to QM9 labels or to the conditions being predicted. The conditional-generation protocol follows Hoogeboom et al. (2022) with the generator and the property classifier trained on disjoint halves of the QM9 split, and the classifier is a pre-existing EGNN; reported MAE values are therefore obtained on structures not seen by the generator, not reproduced from training targets. The SE(3)-invariance claim is supported by a direct geometric proof in Appendix B from the definitions of d, theta, and phi, without importing any uniqueness theorem or self-citation. The GPT-2 vocabulary expansion (Eqs. 12-14) is a construction rather than an assumption, and no parameter fitted to a subset of data is renamed as a prediction of that same subset. The only notable weakness is non-circular: Appendix A's Gram-Schmidt frame (Eqs. 17-18) is undefined for collinear reference atoms, so the tokenizer may have an unstated domain restriction; this is a correctness concern about degeneracy handling, not a reduction of the paper's claims to its inputs.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central method relies on a hand-chosen codebook size, latent dimension, neighbor window, and decoding hyperparameters. The key geometric axiom, non-collinearity of reference atoms, is unstated and load-bearing for the descriptor construction. No fundamentally new physical entities are introduced.

free parameters (5)
  • Codebook size K = 256
    Chosen by hand as a trade-off between reconstruction error and downstream language-modeling difficulty (Section 4.1).
  • Latent embedding dimension = 5
    Chosen by hand for the VQ-VAE encoder output (Section 4.1).
  • Commitment coefficient beta = not stated in text
    Appears in Eq. 11; its value is not given in the paper, though it is a standard VQ-VAE hyperparameter.
  • Number of neighbor atoms for understanding descriptors = 4
    The understanding descriptor ui uses the four nearest neighbors (Section 3.3); this is a hand-chosen window size.
  • Decoding temperature and top-k = tau=0.7, top-k=50
    Selected by hand for the main generation runs; the temperature is analyzed as a diversity/validity trade-off in Section 5.
assumptions (4)
  • domain assumption For every atom, the selected reference atoms f, c1, c2 are non-collinear, so the Gram-Schmidt construction yields a valid orthonormal basis.
    Appendix A defines e2 by normalizing x_c2f - (x_c2f · e1)e1; if the vectors are parallel, e2 is undefined. The paper does not address linear or degenerate geometries.
  • domain assumption The 2D topology-based reference selection (Eq. 5) always produces a unique, well-ordered triple f, c1, c2 for every atom except the first few in the sequence.
    Section 3.2 assumes each atom (beyond the first three) has a preceding topological neighbor chain sufficient to define the frame; no fallback is described for atoms with fewer than three preceding neighbors.
  • domain assumption The VQ-VAE reconstruction error is small enough that the downstream GPT-2 generation and Graphormer understanding tasks are not significantly harmed.
    The paper evaluates reconstruction indirectly through generation validity and property prediction, but does not report a direct task-independent error threshold.
  • standard math Standard properties of rotations and unitary matrices used in the SE(3) invariance proof.
    Appendix B uses the fact that rotation matrices preserve dot products and inner products, which is standard linear algebra.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Tokenizing 3D Molecule Structure with Quantized Spherical Coordinates." pith.science (2026). https://pith.science/paper/NHBJILOL

@misc{pith2026241201564,
  author       = {Pith},
  title        = {Pith review of: Tokenizing 3D Molecule Structure with Quantized Spherical Coordinates},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NHBJILOL}},
  note         = {Machine review of arXiv:2412.01564}
}
read the original abstract

The application of language models (LMs) to molecular structure generation using line notations such as SMILES and SELFIES has been well-established in the field of cheminformatics. However, extending these models to generate 3D molecular structures presents significant challenges. Two primary obstacles emerge: (1) the difficulty in designing a 3D line notation that ensures SE(3)-invariant atomic coordinates, and (2) the non-trivial task of tokenizing continuous coordinates for use in LMs, which inherently require discrete inputs. To address these challenges, we propose Mol-StrucTok, a novel method for tokenizing 3D molecular structures. Our approach comprises two key innovations: (1) We design a line notation for 3D molecules by extracting local atomic coordinates in a spherical coordinate system. This notation builds upon existing 2D line notations and remains agnostic to their specific forms, ensuring compatibility with various molecular representation schemes. (2) We employ a Vector Quantized Variational Autoencoder (VQ-VAE) to tokenize these coordinates, treating them as generation descriptors. To further enhance the representation, we incorporate neighborhood bond lengths and bond angles as understanding descriptors. Leveraging this tokenization framework, we train a GPT-2 style model for 3D molecular generation tasks. Results demonstrate strong performance with significantly faster generation speeds and competitive chemical stability compared to previous methods. Further, by integrating our learned discrete representations into Graphormer model for property prediction on QM9 dataset, Mol-StrucTok reveals consistent improvements across various molecular properties, underscoring the versatility and robustness of our approach.

Figures

Figures reproduced from arXiv: 2412.01564 by the authors.

Figure 1
Figure 1. Converting continuous molecular structures into discrete token sequences. The utilization of language models (LMs) for molec￾ular generation has gained significant traction ow￾ing to their demonstrated success across various tasks (Irwin et al., 2022; Frey et al., 2023; Livne et al., 2024). Typically, molecules are represented as one-dimensional (1D) text strings using different line notation methods, such as SMILES… view at source ↗
Figure 2
Figure 2. (a) Overview of the Mol-StrucTok pipeline. Taking SMILES notation as an example, [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Three types of reference atom selection. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Analysis of structural alphabet and temperature. [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: Decoding generation descriptor distributions. C IMPLEMENTATION DETAILS C.1 VECTOR QUANTIZATION After obtaining the descriptors, we applied normalization to all of them. For the length descriptors, we used log normalization, while for the angle descriptors, we normalize…
Figure 6
Figure 6. Figure 6: Heatmap of atom type hit ratios across shared alphabets. The darker the color, the higher the occurrence probability (log P). RMSD Gen-Length Gen-Polar Gen-Azimuth Mol-StrucTok 0.01 0.06 0.06 - w.o. feature normalization 0.03 0.12 0.11 - w.o. sign prediction 0.01 0.08 …

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Atomic Design Transformer: Scaffold-Conditioned 3D Molecule Generation via xTB-Reward Reinforcement Learning

    physics.comp-ph 2026-07 conditional novelty 6.0 of 10

    A plain causal transformer that tokenizes atom positions in local frames generates 3D molecules directly; RL against an xTB relaxation reward lifts topology-preserving valid yield from ~50% to ~95%.

Reference graph

Works this paper leans on

46 extracted references · 29 canonical work pages · cited by 1 Pith paper

  1. [1]

    PoseBusters : AI -based docking methods fail to generate physically valid poses or generalise to novel sequences

    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 0 (9): 0 3130--3139, 2024

    work page 2024

  2. [2]

    Symphony: S ymmetry-equivariant point-centered spherical harmonics for 3D molecule generation

    Ameya Daigavane, Song Eun Kim, Mario Geiger, and Tess Smidt. Symphony: S ymmetry-equivariant point-centered spherical harmonics for 3D molecule generation. In The Twelfth International Conference on Learning Representations, 2023

    work page 2023

  3. [3]

    Neural scaling of deep chemical models

    Nathan C Frey, Ryan Soklaski, Simon Axelrod, Siddharth Samsi, Rafael Gomez-Bombarelli, Connor W Coley, and Vijay Gadepally. Neural scaling of deep chemical models. Nature Machine Intelligence, 5 0 (11): 0 1297--1305, 2023

    2023

  4. [4]

    Zhangyang Gao, Cheng Tan, Jue Wang, Yufei Huang, Lirong Wu, and Stan Z. Li. FoldToken : Learning protein language via vector quantization and beyond. arXiv preprint arXiv:2403.09673, 2024

    work page Pith review arXiv 2024

  5. [5]

    E (n) equivariant normalizing flows

    Victor Garcia Satorras, Emiel Hoogeboom, Fabian Fuchs, Ingmar Posner, and Max Welling. E (n) equivariant normalizing flows. Advances in Neural Information Processing Systems, 34: 0 4181--4192, 2021

    2021

  6. [6]

    Symmetry-adapted generation of 3D point sets for the targeted discovery of molecules

    Niklas Gebauer, Michael Gastegger, and Kristof Sch \"u tt. Symmetry-adapted generation of 3D point sets for the targeted discovery of molecules. Advances in Neural Information Processing Systems, 32, 2019

    work page 2019

  7. [7]

    Neural message passing for quantum chemistry

    Justin Gilmer, Samuel S Schoenholz, Patrick F Riley, Oriol Vinyals, and George E Dahl. Neural message passing for quantum chemistry. In International Conference on Machine Learning, pp.\ 1263--1272, 2017

    work page 2017

  8. [8]

    xVal : A continuous number encoding for large language models

    Siavash Golkar, Mariel Pettee, Michael Eickenberg, Alberto Bietti, Miles Cranmer, Geraud Krawezik, Francois Lanusse, Michael McCabe, Ruben Ohana, Liam Parker, et al. xVal : A continuous number encoding for large language models. arXiv preprint arXiv:2310.02989, 2023

    arXiv

Show all 46 references
  1. [9]

    Simulating 500 million years of evolution with a language model

    Tomas Hayes, Roshan Rao, Halil Akin, Nicholas J Sofroniew, Deniz Oktay, Zeming Lin, Robert Verkuil, Vincent Q Tran, Jonathan Deaton, Marius Wiggert, et al. Simulating 500 million years of evolution with a language model. bioRxiv, pp.\ 2024--07, 2024

    2024

  2. [10]

    Bilingual language model for protein sequence and structure

    Michael Heinzinger, Konstantin Weissenow, Joaquin Gomez Sanchez, Adrian Henkel, Milot Mirdita, Martin Steinegger, and Burkhard Rost. Bilingual language model for protein sequence and structure. bioRxiv, pp.\ 2023--07, 2023

    2023

  3. [11]

    Equivariant diffusion for molecule generation in 3D

    Emiel Hoogeboom, V ctor Garcia Satorras, Cl \'e ment Vignac, and Max Welling. Equivariant diffusion for molecule generation in 3D . In International Conference on Machine Learning, pp.\ 8867--8887, 2022

    2022

  4. [12]

    Chemformer: a pre-trained transformer for computational chemistry

    Ross Irwin, Spyridon Dimitriadis, Jiazhen He, and Esben Jannik Bjerrum. Chemformer: a pre-trained transformer for computational chemistry. Machine Learning: Science and Technology, 3 0 (1): 0 015022, 2022

    2022

  5. [13]

    Universal structure conversion method for organic molecules: F rom atomic connectivity to three-dimensional geometry

    Yeonjoon Kim and Woo Youn Kim. Universal structure conversion method for organic molecules: F rom atomic connectivity to three-dimensional geometry. Bulletin of the Korean Chemical Society, 36 0 (7): 0 1769--1777, 2015

    2015

  6. [14]

    Self-referencing embedded strings ( SELFIES ): A 100\ Machine Learning: Science and Technology, 1 0 (4): 0 045024, 2020

    Mario Krenn, Florian H \"a se, AkshatKumar Nigam, Pascal Friederich, and Alan Aspuru-Guzik. Self-referencing embedded strings ( SELFIES ): A 100\ Machine Learning: Science and Technology, 1 0 (4): 0 045024, 2020

    2020

  7. [15]

    Geometry informed tokenization of molecules for language model generation

    Xiner Li, Limei Wang, Youzhi Luo, Carl Edwards, Shurui Gui, Yuchao Lin, Heng Ji, and Shuiwang Ji. Geometry informed tokenization of molecules for language model generation. arXiv preprint arXiv:2408.10120, 2024

    2024 arXiv

  8. [16]

    Nach0 : Multimodal natural and chemical languages foundation model

    Micha Livne, Zulfat Miftahutdinov, Elena Tutubalina, Maksim Kuznetsov, Daniil Polykovskiy, Annika Brundyn, Aastha Jhunjhunwala, Anthony Costa, Alex Aliper, Al \'a n Aspuru-Guzik, et al. Nach0 : Multimodal natural and chemical languages foundation model. Chemical Science, 15 0 ...

    2024

  9. [17]

    One transformer can understand both 2D & 3D molecular data

    Shengjie Luo, Tianlang Chen, Yixian Xu, Shuxin Zheng, Tie-Yan Liu, Liwei Wang, and Di He. One transformer can understand both 2D & 3D molecular data. In The Eleventh International Conference on Learning Representations, 2022

    2022

  10. [18]

    An autoregressive flow model for 3D molecular geometry generation from scratch

    Youzhi Luo and Shuiwang Ji. An autoregressive flow model for 3D molecular geometry generation from scratch. In International Conference on Learning Representations, 2022

    2022

  11. [19]

    GraphDF : A discrete flow model for molecular graph generation

    Youzhi Luo, Keqiang Yan, and Shuiwang Ji. GraphDF : A discrete flow model for molecular graph generation. In International Conference on Machine Learning, pp.\ 7192--7203, 2021

    2021

  12. [20]

    PubChemQC project: A large-scale first-principles electronic structure database for data-driven chemistry

    Maho Nakata and Tomomi Shimazaki. PubChemQC project: A large-scale first-principles electronic structure database for data-driven chemistry. Journal of Chemical Information and Modeling, 57 0 (6): 0 1300--1308, 2017

    2017

  13. [21]

    3DMolNet : A generative network for molecular structures

    Vitali Nesterov, Mario Wieser, and Volker Roth. 3DMolNet : A generative network for molecular structures. arXiv preprint arXiv:2010.06477, 2020

    2010 arXiv

  14. [22]

    Open Babel : A n open chemical toolbox

    Noel M O'Boyle, Michael Banck, Craig A James, Chris Morley, Tim Vandermeersch, and Geoffrey R Hutchison. Open Babel : A n open chemical toolbox. Journal of Cheminformatics, 3: 0 1--14, 2011

    2011

  15. [23]

    MolecularRNN : G enerating realistic molecular graphs with optimized properties

    Mariya Popova, Mykhailo Shvets, Junier Oliva, and Olexandr Isayev. MolecularRNN : G enerating realistic molecular graphs with optimized properties. arXiv preprint arXiv:1905.13372, 2019

    1905 arXiv

  16. [24]

    Language models are unsupervised multitask learners

    Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al. Language models are unsupervised multitask learners. OpenAI Blog , 1 0 (8): 0 9, 2019

    2019

  17. [25]

    Quantum chemistry structures and properties of 134 kilo molecules

    Raghunathan Ramakrishnan, Pavlo O Dral, Matthias Rupp, and O Anatole Von Lilienfeld. Quantum chemistry structures and properties of 134 kilo molecules. Scientific Data, 1 0 (1): 0 1--7, 2014

    2014

  18. [26]

    Zero-shot text-to-image generation

    Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea Voss, Alec Radford, Mark Chen, and Ilya Sutskever. Zero-shot text-to-image generation. In International Conference on Machine Learning, pp.\ 8821--8831, 2021

    2021

  19. [27]

    Generating diverse high-fidelity images with VQ-VAE-2

    Ali Razavi, Aaron Van den Oord, and Oriol Vinyals. Generating diverse high-fidelity images with VQ-VAE-2 . Advances in Neural Information Processing Systems, 32, 2019

    2019

  20. [28]

    E (n) equivariant graph neural networks

    V ctor Garcia Satorras, Emiel Hoogeboom, and Max Welling. E (n) equivariant graph neural networks. In International Conference on Machine Learning, pp.\ 9323--9332, 2021

    2021

  21. [29]

    u tt, Huziel E Sauceda, P-J Kindermans, Alexandre Tkatchenko, and K-R M \

    Kristof T Sch \"u tt, Huziel E Sauceda, P-J Kindermans, Alexandre Tkatchenko, and K-R M \"u ller. SchNet -- A deep learning architecture for molecules and materials. The Journal of Chemical Physics, 148 0 (24), 2018

    2018

  22. [30]

    GraphAF : A flow-based autoregressive model for molecular graph generation

    Chence Shi, Minkai Xu, Zhaocheng Zhu, Weinan Zhang, Ming Zhang, and Jian Tang. GraphAF : A flow-based autoregressive model for molecular graph generation. In International Conference on Learning Representations, 2020

    2020

  23. [31]

    Reinforcement learning for molecular design guided by quantum mechanics

    Gregor Simm, Robert Pinsler, and Jos \'e Miguel Hern \'a ndez-Lobato. Reinforcement learning for molecular design guided by quantum mechanics. In International Conference on Machine Learning, pp.\ 8959--8969, 2020

    2020

  24. [32]

    Unified generative modeling of 3D molecules via bayesian flow networks

    Yuxuan Song, Jingjing Gong, Yanru Qu, Hao Zhou, Mingyue Zheng, Jingjing Liu, and Wei-Ying Ma. Unified generative modeling of 3D molecules via bayesian flow networks. In The Twelfth International Conference on Learning Representations, 2023

    2023

  25. [33]

    SaProt : Protein language modeling with structure-aware vocabulary

    Jin Su, Chenchen Han, Yuyang Zhou, Junjie Shan, Xibin Zhou, and Fajie Yuan. SaProt : Protein language modeling with structure-aware vocabulary. In The Twelfth International Conference on Learning Representations, 2023

    2023

  26. [34]

    Neural discrete representation learning

    Aaron Van Den Oord, Oriol Vinyals, et al. Neural discrete representation learning. Advances in Neural Information Processing Systems, 30, 2017

    2017

  27. [35]

    Gilchrist, Johannes S \"o ding, and Martin Steinegger

    Michel van Kempen, Stephanie Kim, Charlotte Tumescheit, Milot Mirdita, Jeongjae Lee, Cameron L.M. Gilchrist, Johannes S \"o ding, and Martin Steinegger. Fast and accurate protein structure search with foldseek. Nature Biotechnology, 42: 0 243 -- 246, 2022

    2022

  28. [36]

    Attention is all you need

    A Vaswani. Attention is all you need. Advances in Neural Information Processing Systems, 2017

    2017

  29. [37]

    SMILES , a chemical language and information system

    David Weininger. SMILES , a chemical language and information system. Journal of Chemical Information and Computer Sciences, 28 0 (1): 0 31--36, 1988

    1988

  30. [38]

    Geometric latent diffusion models for 3D molecule generation

    Minkai Xu, Alexander S Powers, Ron O Dror, Stefano Ermon, and Jure Leskovec. Geometric latent diffusion models for 3D molecule generation. In International Conference on Machine Learning, pp.\ 38592--38610, 2023

    2023

  31. [39]

    Do transformers really perform badly for graph representation? Advances in Neural Information Processing Systems, 34: 0 28877--28888, 2021

    Chengxuan Ying, Tianle Cai, Shengjie Luo, Shuxin Zheng, Guolin Ke, Di He, Yanming Shen, and Tie-Yan Liu. Do transformers really perform badly for graph representation? Advances in Neural Information Processing Systems, 34: 0 28877--28888, 2021

    2021

  32. [40]

    Latent 3D graph diffusion

    Yuning You, Ruida Zhou, Jiwoong Park, Haotian Xu, Chao Tian, Zhangyang Wang, and Yang Shen. Latent 3D graph diffusion. In The Twelfth International Conference on Learning Representations, 2023

    2023

  33. [41]

    An equivariant generative framework for molecular graph-structure co-design

    Zaixi Zhang, Qi Liu, Chee-Kong Lee, Chang-Yu Hsieh, and Enhong Chen. An equivariant generative framework for molecular graph-structure co-design. Chemical Science, 14 0 (31): 0 8380--8392, 2023

    2023

  34. [42]

    BindGPT : A scalable framework for 3d molecular design via language modeling and reinforcement learning

    Artem Zholus, Maksim Kuznetsov, Roman Schutski, Rim Shayakhmetov, Daniil Polykovskiy, Sarath Chandar, and Alex Zhavoronkov. BindGPT : A scalable framework for 3d molecular design via language modeling and reinforcement learning. arXiv preprint arXiv:2406.03686, 2024

    2024 arXiv

  35. [43]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

  36. [44]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should ...

  37. [45]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@firs...

  38. [46]

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibset...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.