Liu, Zhi-Song, Peng, Wenqing, Kheder, Ammar, Makkonen, Risto, Boy, Michael · Oct 24, 2026
arXiv cs.AI paper presents a neural operator for near-explicit atmospheric chemistry
Markdown · JSON · Original source
The paper’s authors introduce FNO-, a Fourier neural operator with a mixture of experts, to predict the full near-explicit chemical state over 30 minutes in a single call. They say the model also uses disagreement among its experts to select informative box-model simulations, and that this active sampling produces a more accurate emulator than random or evenly spread sampling at equal cost. The authors report that on longer trajectories keeps the emulator stable over multi-hour runs where models trained on short simulations diverge. They say it reproduces the evolution of thousands of species with high fidelity, handles unseen chemical compositions, and runs orders of magnitude faster than the numerical solver. The paper frames training under changing meteorology as the next step toward bringing near-explicit chemistry into Earth-system models; it does not establish that this step has been achieved.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
cs.AI · Article
Source details
Original source date: Oct 24, 2026 (2026-10-24).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-14T01:33:43.646Z. First observed: 2026-10-28T06:10:32.180Z. Retrieved: 2026-10-30T22:31:36.942Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
The abstract does not provide the numerical measures or evaluation conditions behind its reported accuracy, stability, and speed comparisons. The authors identify training under changing meteorology as future work; the abstract does not report results for that condition.
Open question
How well would FNO- perform when trained under changing meteorology, the next step the authors identify for bringing near-explicit chemistry into Earth-system models?
Source-stated
The authors introduce FNO-MoE, which predicts the full near-explicit chemical state over 30 minutes in a single call.
Here we introduce FNO-MoE, a Fourier neural operator with a mixture of experts that predicts the full near-explicit chemical state over 30 minutes in a single call.
Source-stated
The authors report that active sampling using disagreement among the model’s experts yields a more accurate emulator than random or evenly spread sampling at equal cost.
At equal cost, this active sampling yields a more accurate emulator than random or evenly spread sampling.
Source-stated
The authors report stability over multi-hour runs after fine-tuning on longer trajectories, where models trained on short simulations diverge.
Fine-tuning on longer trajectories keeps it stable over multi-hour runs in which models trained on short simulations diverge.
Source-stated
The authors report high-fidelity reproduction of thousands of species, accuracy for unseen chemical compositions, and much greater speed than the numerical solver.
The resulting emulator reproduces the evolution of thousands of species with high fidelity, remains accurate for unseen chemical compositions and runs orders of magnitude faster than the numerical solver.
Source-stated
The authors identify training under changing meteorology as a next step, not as a demonstrated capability of this work.
Training under changing meteorology is the next step towards bringing near-explicit chemistry into Earth-system models.
Source and contributor context
Liu, Zhi-Song, Peng, Wenqing, Kheder, Ammar, Makkonen, Risto, Boy, Michael · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Mixture of experts
A model design that sends each input to a few specialized sub-networks instead of the whole model, which saves computation.
Fine-tuning
Further training of an existing model to adapt it to a particular task or use.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Editorial provenance
Version 1. Publication ID: 9d05c0a3-8b48-46a6-869a-a2c7d0cf4ccd.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-14T01:33:43.646Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 12d5c5d0cd91cdd14e758c5846e03d157de2288deafca17f7d2ae5eda1659d68. Source text hash: 86e49b6e347a2e733590b00d5f68be4a9ddf54cd674e56973d362a0703ba57ce.
Revision and source-check history
Version 1 · 2026-11-14T01:33:43.646Z
arXiv cs.AI paper presents a neural operator for near-explicit atmospheric chemistry
The paper’s authors introduce FNO-, a Fourier neural operator with a mixture of experts, to predict the full near-explicit chemical state over 30 minutes in a single call. They say the model also uses disagreement among its experts to select informative box-model simulations, and that this active sampling produces a more accurate emulator than random or evenly spread sampling at equal cost. The authors report that on longer trajectories keeps the emulator stable over multi-hour runs where models trained on short simulations diverge. They say it reproduces the evolution of thousands of species with high fidelity, handles unseen chemical compositions, and runs orders of magnitude faster than the numerical solver. The paper frames training under changing meteorology as the next step toward bringing near-explicit chemistry into Earth-system models; it does not establish that this step has been achieved.
Original source
cs.AI · Article
Source details
Original source date: Oct 24, 2026 (2026-10-24).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-14T01:33:43.646Z. First observed: 2026-10-28T06:10:32.180Z. Retrieved: 2026-10-30T22:31:36.942Z.
Limits
The abstract does not provide the numerical measures or evaluation conditions behind its reported accuracy, stability, and speed comparisons. The authors identify training under changing meteorology as future work; the abstract does not report results for that condition.
Open question
How well would FNO- perform when trained under changing meteorology, the next step the authors identify for bringing near-explicit chemistry into Earth-system models?
Source-stated
The authors introduce FNO-MoE, which predicts the full near-explicit chemical state over 30 minutes in a single call.
Here we introduce FNO-MoE, a Fourier neural operator with a mixture of experts that predicts the full near-explicit chemical state over 30 minutes in a single call.
Source-stated
The authors report that active sampling using disagreement among the model’s experts yields a more accurate emulator than random or evenly spread sampling at equal cost.
At equal cost, this active sampling yields a more accurate emulator than random or evenly spread sampling.
Source-stated
The authors report stability over multi-hour runs after fine-tuning on longer trajectories, where models trained on short simulations diverge.
Fine-tuning on longer trajectories keeps it stable over multi-hour runs in which models trained on short simulations diverge.
Source-stated
The authors report high-fidelity reproduction of thousands of species, accuracy for unseen chemical compositions, and much greater speed than the numerical solver.
The resulting emulator reproduces the evolution of thousands of species with high fidelity, remains accurate for unseen chemical compositions and runs orders of magnitude faster than the numerical solver.
Source-stated
The authors identify training under changing meteorology as a next step, not as a demonstrated capability of this work.
Training under changing meteorology is the next step towards bringing near-explicit chemistry into Earth-system models.
Source and contributor context
Liu, Zhi-Song, Peng, Wenqing, Kheder, Ammar, Makkonen, Risto, Boy, Michael · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Mixture of experts
A model design that sends each input to a few specialized sub-networks instead of the whole model, which saves computation.
Fine-tuning
Further training of an existing model to adapt it to a particular task or use.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Editorial provenance
Version 1. Publication ID: 9d05c0a3-8b48-46a6-869a-a2c7d0cf4ccd.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-14T01:33:43.646Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 12d5c5d0cd91cdd14e758c5846e03d157de2288deafca17f7d2ae5eda1659d68. Source text hash: 86e49b6e347a2e733590b00d5f68be4a9ddf54cd674e56973d362a0703ba57ce.
uncertain · 2026-11-14T01:33:43.646Z · anthropic/claude-opus-5.5
Publication: 9d05c0a3-8b48-46a6-869a-a2c7d0cf4ccd
attribution — pass: The reading consistently attributes the claims to the paper's authors, using phrases like 'the authors report' and 'they say'. The authors are listed on the arXiv page, and the reading presents no source assertion as established fact.
Authors:Zhi-Song Liu, Wenqing Peng, Ammar Kheder, Risto Makkonen, Michael Boy
summary — uncertain: The reading leaves out the problem the paper addresses. Earth-system models simplify chemistry because near-explicit mechanisms such as MCM are too costly to solve in every grid cell. Without that, a general reader cannot see why a fast emulator matters. Central terms are also left unexplained: near-explicit chemistry, emulator, box-model simulations, Earth-system models, and Fourier neural operator / mixture of experts (which could be handled with tooltips). Smallest repair: open the summary with a plain-language version of the abstract's first sentence on condensed versus near-explicit mechanisms and cost. The question is substantive and grounded in the source.
Earth-system models simplify atmospheric chemistry into condensed mechanisms, because near-explicit mechanisms such as the Master Chemical Mechanism (MCM) are too costly to solve in every grid cell.
date — pass: The originalPublished date of 2026-10-24 matches the v1 submission date. It is not the date of the later v2 revision or the feed date.
Submitted on 24 Oct 2026 (v1), last revised 27 Oct 2026 (this version, v2)
support — pass: Each claim is quoted from the abstract and labelled as source-stated. The reading correctly frames training under changing meteorology as future work rather than a demonstrated result. The limitations accurately note that the abstract gives no figures. It adds no unsupported benefits, costs or forecasts.
At equal cost, this active sampling yields a more accurate emulator than random or evenly spread sampling.
media — pass: The selected media is the verified direct link to the original arXiv abstract page. It is the most direct way to reach the source.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
A model design that sends each input to a few specialized sub-networks instead of the whole model, which saves computation.
Definition referenceFurther training of an existing model to adapt it to a particular task or use.
Definition referencearXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivA model design that sends each input to a few specialized sub-networks instead of the whole model, which saves computation.
Definition referenceA model design that sends each input to a few specialized sub-networks instead of the whole model, which saves computation.
Definition referenceFurther training of an existing model to adapt it to a particular task or use.
Definition referencearXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivA model design that sends each input to a few specialized sub-networks instead of the whole model, which saves computation.
Definition reference