Pith. sign in

REVIEW 2 major objections 5 minor 33 references

Chemical filters for ultra-high-throughput materials screening and generation

T0 review · 2 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read A tunable oxidation-state filter can screen and guide AI-generated materials, preserving low-energy crystals while rejecting chemically implausible ones.

desk verdict Useful, well-scoped methods paper with a real soft spot in the MLIP-based stability validation; the tunable filter itself is solid and worth refereeing. read the letter →

arxiv 2607.17910 v1 pith:I4TAADXY submitted 2026-07-20 cond-mat.mtrl-sci cs.AIcs.LG

Kinga O. Mastej , Panyalak Detrattanawichai , Hyunsoo Park , Anthony Onwuli , Masahiro Negishi , Aron Walsh This is my paper
classification cond-mat.mtrl-scics.AIcs.LG
keywords chemicalvalidityoxidationstatesgenerativemodelsmaterialsdiscoveryconvexhullreinforcementlearningscreeningcompositionalfiltering
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 argues that chemical plausibility in AI-generated materials is not a binary yes/no but a matter of degree, and that it can be captured by a tunable operator built from oxidation-state statistics in experimental crystal data. The operator has two knobs: a consensus threshold, which drops oxidation states rarely seen in the database, and a commonality threshold, which keeps only states that account for a large enough share of an element's recorded occurrences. With these knobs, one can move continuously between permissive exploration and conservative down-selection. Across six generative models, the filter shows that most models get stoichiometry right but under-sample realistic oxidation-state combinations, and that stricter filtering preferentially removes high-energy compounds while retaining low-energy ones near the convex hull. The same operator can also act as a reinforcement-learning reward, steering a latent diffusion model toward more metastable, chemically grounded compositions.

What carries the argument

The load-bearing component is a composition-level chemical validity operator built from empirical oxidation-state statistics. For each element, a filtered set of allowed oxidation states is constructed by keeping states that appear at least C times in a large experimental crystal-structure database (consensus) and that account for at least a fraction θ of that element's recorded occurrences (commonality). A composition is declared valid if some product assignment from these per-element sets is charge neutral and satisfies the Pauling electronegativity rule. For the continuous reward, the same machinery is turned into a score: the geometric mean, over the elements in a composition, of the obs

What would settle it

Compile a list of experimentally confirmed inorganic compounds that rely on oxidation states with fewer than three database occurrences or with low relative frequency (for example, certain high-valence oxides or stabilized low-valence species). If the strict filter (C=3, θ=0.5) rejects a large share of these confirmed compounds, that is direct evidence that the database prior is too conservative; conversely, if it retains them, the prior tracks synthesizability closely.

Watch

Extended reading notes

Core claim

The central claim is that oxidation-state frequencies observed in experimental crystal structures provide a model-agnostic chemical prior that can both evaluate and guide generative models. The paper formalizes this as a validity operator that accepts only compositions for which some assignment of element oxidation states is charge neutral and obeys the electronegativity rule, where the allowed states are those that survive two adjustable thresholds: consensus C (a minimum absolute number of database occurrences) and commonality θ (a minimum share of an element's occurrences). Benchmarks on six generative models show that while 93–97% of generated compositions pass charge neutrality, strict

Load-bearing premise

The entire screening and reward signal is only as good as the oxidation-state frequencies in the experimental crystal-structure database, so if that database systematically under-records rare but synthesizable oxidation states, the filter will discard legitimate chemistry.

Editorial extensions

If this is right

  • Generative models that score well on standard validity checks can still be far from the empirical oxidation-state distribution; the stricter commonality filter exposes this gap, with retention dropping to 42–55% at θ=0.5.
  • Charge neutrality and electronegativity consistency are not enough to predict stability; empirical oxidation-state commonality adds a chemical signal that correlates with low energy above the convex hull.
  • Stricter filtering preferentially removes high-Ehull compounds, so the operator acts as a cheap first-pass stability prior before expensive structure relaxation.
  • The same operator can be used as an in-training reward without retraining the generative architecture, since it is model-agnostic.
  • Oxidation-state-aware rewards can trade novelty and diversity for metastability; the continuous reward raised metastability while lowering uniqueness and novelty.

Reading between the lines

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

  • Because the operator is tunable, it could be adapted to synthesis-specific goals: loosening θ to propose compounds that require unusual stabilisation, or tightening it near the end of a pipeline to favor well-trodden chemistry.
  • The correlation between validity and Ehull may reflect that common oxidation states are often those with accessible synthesis routes; if so, the filter indirectly encodes kinetic or thermodynamic accessibility beyond what the convex-hull calculation captures. This connection is not tested in the paper.
  • A testable extension: use the continuous reward in a multi-objective setup with an exploration bonus, which could recover some of the lost novelty while retaining the metastability gain.
  • As the underlying database grows and updates, the same operator can be re-derived without changing the method; quantifying how much rare-but-real chemistry is discarded at each threshold remains an open empirical question.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The paper introduces a tunable chemical validity operator built on the SMACT oxidation-state model, with a consensus threshold C and a commonality threshold θ that control the strictness of empirical oxidation-state prior. The authors benchmark six generative models (ADiT, CDVAE, Chemeleon-DNG, DiffCSP, DiffCSP++, MatterGen) on 10,000 generated compositions each, showing that charge neutrality and the electronegativity rule are usually satisfied, while agreement with empirical oxidation-state commonality varies substantially across models. They further analyze energy above the convex hull (Ehull) using MACE-MPA-0 relaxations, reporting that stricter validity filtering correlates with lower Ehull. Finally, they demonstrate the operator as a binary or continuous RL reward for fine-tuning the Chemeleon2 latent diffusion model, claiming the continuous reward improves metastability and mSUN at the cost of uniqueness and novelty.

Significance. If the claims hold, this work provides a simple, interpretable, and model-agnostic chemical filter that can serve both as a post-hoc diagnostic and as an in-training reward for generative models. The threshold formulation is a pragmatic extension of SMACT, and the open-source code and benchmark data are valuable resources. The benchmarking across six state-of-the-art models is a useful contribution. However, the central Ehull-based evidence rests on MLIP energies with potentially systematic bias, and the RL results are unreplicated, so the quantitative claims require strengthening before the significance can be fully assessed.

major comments (2)
  1. [§4.4 / Fig. 2] The claim that stricter filtering preserves low-energy compounds is based on the Ehull comparison in Fig. 2, computed with MACE-MPA-0, which the paper states has errors of 'a few tens of meV/atom' (§4.4). At C=3, θ=0.5 the median valid–invalid gap is 0.089 eV/atom (89 meV/atom), the same order of magnitude as the reported MLIP error. Since MACE-MPA-0 is trained on Materials Project structures dominated by common oxidation states, energies for compounds with rare oxidation states may be systematically biased, inflating the observed separation. The paper does not test for systematic bias with respect to the validity grouping variable. I recommend a DFT cross-check on a stratified sample (e.g., 50–100 valid and invalid compounds balanced across oxidation-state commonality) or, at minimum, an analysis of residual errors against oxidation-state frequency. Without this, the 'preserves low-ener
  2. [§2.4 / Fig. 6] The reinforcement-learning results are presented from single training runs. Figure 6 shows one reward trajectory and one set of final metrics per reward, with no error bars or multiple seeds. The observed changes (e.g., mSUN 0.0764→0.0908, metastability 0.3279→0.4310) could be within run-to-run variance, especially with only 400 update steps and stochastic generation. The claims that the continuous SMACT reward improves metastability and mSUN, and that the binary reward does not, require reproducibility evidence. Please report means and standard deviations over at least 3–5 independent seeds, or otherwise justify the stability of the results.
minor comments (5)
  1. [§2.2 / Table 1] The sentence 'reducing the fraction of retained compositions by a further 3–7%' is inconsistent with Table 1: from q+χ to C=3 the decreases are 1.7%, 4.2%, 3.7%, 4.0%, 3.8%, 3.9% for ADiT, CDVAE, Chemeleon-DNG, DiffCSP, DiffCSP++, and MatterGen. Similarly, the '42–55%' range for θ=0.5 omits CDVAE's 41.8% (or should be rounded consistently). Please correct the text.
  2. [§2.4 / Fig. 6] The term 'mSUN' is used in the Results but only defined in the Methods. Please define it at first use in the Results. Also, specify the number of independent runs/seeds in the figure caption.
  3. [§4.1 / Eq. (2)] Clarify how include_zero=False interacts with the commonality calculation: is the denominator in Eq. (2) computed over the zero-inclusive or zero-excluded oxidation-state set? This could affect the filtered Ω set and the resulting retention rates.
  4. [§4.4 / Fig. 2] The exclusion of Yb-containing compositions (2.5%) is stated only in the Methods. Please mention this important data exclusion in the Results text for Fig. 2, as it may affect the comparability of the Ehull distributions.
  5. [§2.3 / Fig. 3] The caption of Fig. 3 should state explicitly which oxidation-state list and which consensus setting (C=1 vs C=3) are used for the 'permissive' and 'strictest' panels. The current text leaves some ambiguity about the default parameters.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: validity labels and stability/energy metrics are constructed independently, and the load-bearing claims are empirical correlations rather than definitional reductions.

full rationale

The derivation chain is self-contained on inspection. The oxidation-state prior is explicitly constructed from ICSD frequencies in Eqs. (1)-(3), and the validity operator in Eqs. (4)-(7) is a stated composition-level rule (charge neutrality, electronegativity ordering, membership in the filtered state set); nothing in this construction is fitted to or defined by the downstream quantities it is used to evaluate. The central quantitative claim, that stricter filtering retains low-Ehull compounds while invalid compositions concentrate in the high-Ehull tail (Figure 2), compares these validity labels against energies above the convex hull computed with MACE-MPA-0 (Methods 4.4), an independent surrogate for DFT stability that does not appear in the validity definition. The RL reward in Eq. (8) uses the same ICSD oxidation-state proportions, but the reported gains in mSUN/metastability are evaluated with external Ehull and diversity metrics; the paper even shows a metric-targeted reward outperforms on aligned metrics, indicating the authors are not conflating reward and evaluation. Retention curves in Figures 3-5 are descriptive consequences of the threshold definitions, not predictions, and are presented as tunability demonstrations. Self-citations to SMACT and Chemeleon identify codebases/models whose relevant equations and settings are stated in the manuscript; no load-bearing argument reduces to an unverified self-citation. Limitations on ICSD coverage bias and MLIP energy error are acknowledged in the Discussion and Methods 4.4; they are accuracy concerns, not evidence of circularity.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim depends on user-set thresholds C and θ, on ICSD data as ground truth, on Pauling's rule, and on MACE-MPA-0 energies for the Ehull correlation. No new physical entities are introduced.

free parameters (2)
  • Consensus threshold C = C=3 (default); also C=1 and C=0 used
    User-set parameter defining minimum number of ICSD occurrences for an oxidation state to be allowed (Eq. 1). Determines the strictness of the oxidation-state set.
  • Commonality threshold θ = θ=0.05, 0.10, 0.50
    User-set parameter defining minimum fraction of an element's recorded oxidation-state occurrences (Eq. 2). Controls retention from permissive to conservative.
assumptions (4)
  • domain assumption ICSD oxidation-state frequencies are a reliable empirical prior for chemical plausibility.
    All filtering uses f(E^q) from ICSD (Eqs. 1-2); if this data is biased, the filter is biased.
  • domain assumption The Pauling electronegativity rule is a valid necessary condition for charge assignment.
    Used as a hard constraint in Eq. (6); not derived or validated in this paper.
  • domain assumption MACE-MPA-0 foundation potential predicts energies accurately enough for distributional comparisons of Ehull.
    Used to relax ~60k structures and compute convex hull distances (Section 4.4); the paper acknowledges tens of meV errors.
  • domain assumption The 0.1 eV/atom threshold defines metastability.
    Standard convention from Sun et al. [27], used to classify metastable vs unstable.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Chemical filters for ultra-high-throughput materials screening and generation." pith.science (2026). https://pith.science/paper/I4TAADXY

@misc{pith2026260717910,
  author       = {Pith},
  title        = {Pith review of: Chemical filters for ultra-high-throughput materials screening and generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/I4TAADXY}},
  note         = {Machine review of arXiv:2607.17910}
}
read the original abstract

Generative artificial intelligence is rapidly transforming materials design by enabling de novo exploration of immense chemical spaces. Yet a large proportion of AI-generated compositions remain implausible, violating established chemical principles, which limits the reliability and interpretability of generative materials design. Here, we introduce a chemical validity operator that recasts heuristic chemical rules as a configurable algorithmic prior for evaluating and guiding generative materials discovery. Built on the open-source SMACT package, a data-informed oxidation-state model exposes tunable thresholds, allowing users to interpolate continuously between permissive and conservative chemical constraints, while supporting both exploratory and conservative materials-design workflows. Benchmarking six state-of-the-art generative models for inorganic crystals shows that most reproduce stoichiometry but under-represent realistic oxidation-state combinations, and that filtering removes compositions reliant on rarely observed oxidation states while preserving low-energy compounds near the convex hull. Beyond screening, the same operator can also serve as a reinforcement-learning reward, steering a latent diffusion model towards chemically grounded compositions. By encoding chemical heuristics and observations, this work establishes a foundation for oxidation-state-aware generative models.

Figures

Figures reproduced from arXiv: 2607.17910 by the authors.

Figure 1
Figure 1. Oxidation state distributions across the periodic table. (a) Representative manganese [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Distributions of energy above the convex hull ( [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Number of oxidation states retained for each element under four filtering settings. The [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Retention of Materials Project and ICSD entries as a function of the oxidation state [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Pruning of the ABX3 perovskite compositional design space under progressively stricter SMACT filtering settings. Each panel shows admissible A and B element combinations for oxide and halide perovskites with X = O, Cl, Br, or I. Blue markers denote compositions satisfy…
Figure 6
Figure 6. Figure 6: Reinforcement learning of Chemeleon2 with chemically informed rewards. (a) Training reward trajectories for binary and continuous SMACT rewards. The binary reward saturates rapidly, whereas the continuous reward increases more gradually because it retains information a…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

33 extracted references · 3 linked inside Pith

  1. [1]

    Zeni, C.et al.A generative model for inorganic materials design.Nature639, 624–632 (2025)

    2025

  2. [2]

    Handoko, A. D. & Made, R. I. Artificial intelligence and generative models for materials discovery – a review. Preprint athttps://arxiv.org/abs/2508.03278(2025)

    arXiv 2025

  3. [3]

    Merchant, A.et al.Scaling deep learning for materials discovery.Nature624, 80–85 (2023)

    2023

  4. [4]

    Preprint athttps://arxiv.org/abs/2512.04562(2025)

    Betala, S.et al.LeMat-GenBench: a unified evaluation framework for crystal generative models. Preprint athttps://arxiv.org/abs/2512.04562(2025). 15

    arXiv 2025

  5. [5]

    Achary, S. N. & Tyagi, A. K. Synthesis of materials with unusual oxidation state. in Handbook on Synthesis Strategies for Advanced Materials: Volume-III: Materials Specific Synthesis Strategies(eds Tyagi, A. K. & Ningthoujam, R. S.) 137–158 (Springer, 2021)

    2021

  6. [6]

    npj Comput

    Deng, B.et al.Systematic softening in universal machine learning interatomic potentials. npj Comput. Mater.11, 9 (2025)

    2025

  7. [7]

    & Zavadlav, J

    Sanocki, M. & Zavadlav, J. Generalization of long-range machine learning potentials in complex chemical spaces. Preprint athttps://arxiv.org/abs/2512.10989(2025)

    arXiv 2025

  8. [8]

    Goldschmidt, V. M. Die Gesetze der Krystallochemie.Naturwissenschaften14, 477–485 (1926)

    1926

Show all 33 references
  1. [9]

    Pettifor, D. G. A chemical scale for crystal-structure maps.Solid State Commun.51, 31–34 (1984)

    1984

  2. [10]

    Pamplin, B. R. A systematic method of deriving new semiconducting compounds by struc- tural analogy.J. Phys. Chem. Solids25, 675–684 (1964)

    1964

  3. [11]

    Preprint athttps://arxiv.org/abs/2507.19799(2025)

    Cheng, M.et al.Enhancing materials discovery with valence constrained design in genera- tive modeling. Preprint athttps://arxiv.org/abs/2507.19799(2025)

    2025 arXiv

  4. [12]

    Mater.9, 38 (2023)

    Zhao, Y.et al.Physics guided deep learning for generative design of crystal materials with symmetry constraints.npj Comput. Mater.9, 38 (2023)

    2023

  5. [13]

    W.et al.Computational screening of all stoichiometric inorganic materials

    Davies, D. W.et al.Computational screening of all stoichiometric inorganic materials. Chem1, 617–627 (2016)

    2016

  6. [14]

    W., Butler, K

    Davies, D. W., Butler, K. T., Isayev, O. & Walsh, A. Materials discovery by chemical analogy: role of oxidation states in structure prediction.Faraday Discuss.211, 553–568 (2018)

    2018

  7. [15]

    W.et al.SMACT: semiconducting materials by analogy and chemical theory

    Davies, D. W.et al.SMACT: semiconducting materials by analogy and chemical theory. J. Open Source Softw.4, 1361 (2019)

    2019

  8. [16]

    & Jaakkola, T

    Xie, T., Fu, X., Ganea, O.-E., Barzilay, R. & Jaakkola, T. S. Crystal diffusion variational autoencoder for periodic material generation. inInternational Conference on Learning Representations(2022)

    2022

  9. [17]

    & Walsh, A

    Park, H. & Walsh, A. Guiding generative models to uncover diverse and novel crystals via reinforcement learning. Preprint athttps://arxiv.org/abs/2511.07158(2025)

    2025

  10. [18]

    Mater.11, 184 (2025)

    Qiu, Z.et al.Massive discovery of crystal structures across dimensionalities by leveraging vector quantization.npj Comput. Mater.11, 184 (2025)

    2025

  11. [19]

    S., Hanisch, M

    Yan, L., Kang, B. S., Hanisch, M. D., Ma, J. & Anandkumar, A. MGB: the material gen- eration benchmark. Preprint athttps://openreview.net/forum?id=K15Dqxm0ge(2025)

    2025

  12. [20]

    Sci.11, 2304305 (2024)

    Wei, L.et al.Crystal composition transformer: self-learning neural language model for generative and tinkering design of materials.Adv. Sci.11, 2304305 (2024)

    2024

  13. [21]

    K.et al.All-atom diffusion transformers: unified generative modelling of molecules and materials

    Joshi, C. K.et al.All-atom diffusion transformers: unified generative modelling of molecules and materials. inInternational Conference on Machine Learning(2025)

    2025

  14. [22]

    & Walsh, A

    Park, H., Onwuli, A. & Walsh, A. Exploration of crystal chemical space using text-guided generative artificial intelligence.Nat. Commun.16, 4379 (2025). 16

    2025

  15. [23]

    inAdvances in Neural Information Processing Systems, vol

    Jiao, R.et al.Crystal structure prediction by joint equivariant diffusion. inAdvances in Neural Information Processing Systems, vol. 36, 17464–17497 (2023)

    2023

  16. [24]

    Spacegroupconstrainedcrystalgeneration

    Jiao, R., Huang, W., Liu, Y., Zhao, D.&Liu, Y. Spacegroupconstrainedcrystalgeneration. inInternational Conference on Learning Representations(2024)

    2024

  17. [25]

    & Rehme, S

    Zagorac, D., Müller, H., Ruehl, S., Zagorac, J. & Rehme, S. Recent developments in the Inorganic Crystal Structure Database: theoretical crystal structure data and related features.J. Appl. Crystallogr.52, 918–925 (2019)

    2019

  18. [26]

    Jain, A.et al.Commentary: The Materials Project: a materials genome approach to accelerating materials innovation.APL Mater.1, 011002 (2013)

    2013

  19. [27]

    Adv.2, e1600225 (2016)

    Sun, W.et al.The thermodynamic scale of inorganic crystalline metastability.Sci. Adv.2, e1600225 (2016)

    2016

  20. [28]

    Batatia, I.et al.A foundation model for atomistic materials chemistry.J. Chem. Phys. 163, 184110 (2025)

    2025

  21. [29]

    Cohen, O.et al.TorchSim: an efficient atomistic simulation engine in PyTorch.AI Sci.1, 025003 (2025)

    2025

  22. [30]

    Riebesell, J.et al.A framework to evaluate machine learning crystal stability predictions. Nat. Mach. Intell.7, 836–847 (2025)

    2025

  23. [31]

    Preprint athttps://arxiv.org/abs/2402.03300(2024)

    Shao, Z.et al.DeepSeekMath: pushing the limits of mathematical reasoning in open language models. Preprint athttps://arxiv.org/abs/2402.03300(2024)

    2024 arXiv

  24. [32]

    & Ermon, S

    Song, J., Meng, C. & Ermon, S. Denoising diffusion implicit models. inInternational Conference on Learning Representations(2021)

    2021

  25. [33]

    & Hutter, F

    Loshchilov, I. & Hutter, F. Decoupled weight decay regularization. inInternational Con- ference on Learning Representations(2019). 17 Supplementary Information Figure S1: Number of oxidation states listed per element in each of the five built-in oxidation- state lists. Each li...

    2019

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.