Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Large Language Models to Accelerate Organic Chemistry Synthesis

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read The paper reports that a chemistry-tuned language model, trained on 1.28 million reaction Q&A pairs, outperforms prior methods on retrosynthesis and yield prediction and, in a 15-run human-AI campaign, found a 67%-yield condition for a…

desk verdict Worth refereeing: the benchmark and Chemma-BO results are solid, but the 'autonomous open-space' claim is undercut by the chemists' solvent switch and the paper's own Limitations. read the letter →

arxiv 2504.18340 v1 pith:4KGPRHB2 submitted 2025-04-25 physics.chem-ph

Yu Zhang , Yang Han , Shuai Chen , Ruijie Yu , Xin Zhao , Xianbin Liu , Kaipeng Zeng , Mengdi Yu
show 5 more authors
Jidong Tian Feng Zhu Xiaokang Yang Yaohui Jin Yanyan Xu
This is my paper · ORCID
classification physics.chem-ph
keywords organicsynthesislargelanguagemodelsretrosynthesisyieldpredictionactivelearningBayesianoptimizationSuzuki-MiyauracouplingSMILES
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

Chemma is a large language model built by fully fine-tuning a 7-billion-parameter decoder-only model on 1.28 million question-answer pairs about organic reactions, expressed as SMILES and natural language. The paper sets out to show that such a chemistry-tuned model can act as a synthesis assistant: it reports 72.2% top-1 accuracy on single-step retrosynthesis, yield and selectivity predictions with R2 values of 0.74–0.93 on several benchmark and high-throughput datasets, and faster reaction optimization when combined with Bayesian optimization and an active-learning loop. The decisive demonstration is a previously unreported Suzuki-Miyaura coupling of cyclic aminoboronates with aryl halides, where the model and chemists together found a working ligand and solvent in 15 wet-lab runs, reaching 67% isolated yield. If the claims hold, reaction knowledge can be extracted from text and SMILES sequences rather than from quantum-chemical calculations, lowering the barrier to AI-guided synthesis.

What carries the argument

The load-bearing mechanism is Chemma, with three linked components. Chemma-SFT is a decoder-only language model fully fine-tuned on 1.28 million reaction Q&A pairs for forward prediction, retrosynthesis, and condition generation. Chemma-RM is trained with pairwise ranking data and reinforcement learning from experimental feedback, so it prefers conditions associated with higher yields or selectivities. For regression tasks, the authors take reaction embeddings from the last hidden layer of Chemma-SFT and pass them through a small multilayer perceptron that predicts yield or selectivity; the same embedding space feeds the active-learning loop and a modified Bayesian-optimization scheme in which a Gaussian process corrects the bias between Chemma-predicted and observed yields. What carries the argument is that all tasks share one learned representation of reactions, so knowledge acquired from large reaction corpora can be reused for wet-lab suggestions.

What would settle it

Run the same unreported coupling from the same round-0 starting conditions, but replace Chemma's ligand suggestions with random ligands and follow the same human solvent-switch rule; if this control reaches 67% isolated yield in 15 or fewer runs, the attribution of the acceleration to Chemma is falsified. A complementary check: pre-register Chemma's predicted yields for the round-1 candidates and compare them with observed yields; if predictions are not positively correlated with the 67% result, the model is not driving the success.

Watch

Extended reading notes

Core claim

The central claim is that one fully fine-tuned language model, trained only on reaction SMILES paired with instruction-style questions, can learn enough chemistry to outperform dedicated methods on standard predictive tasks and to propose working conditions for reactions outside its training set. On a standard 50,000-reaction retrosynthesis benchmark, Chemma reaches 72.2% top-1 accuracy, surpassing the best transformer baseline by 17.1 points. On high-throughput and literature-derived yield datasets, the model reaches R2 values between 0.74 and 0.86 without DFT descriptors, and on regioselectivity and enantioselectivity tasks it reports R2 values of 0.93 and 0.89. The paper's headline wet-lab result is an unreported Suzuki-Miyaura coupling of a cyclic aminoboronate with an aryl halide: in an active-learning campaign of 15 runs, PAd3 as ligand and 1,4-dioxane as solvent gave 67% isolated yield. The authors argue this shows open-space reaction exploration is possible without predefined condition pools and without quantum-chemical calculations.

Load-bearing premise

The load-bearing premise is that the wet-lab campaign's success—67% isolated yield in 15 runs—can be attributed to Chemma's suggestions rather than to the chemists' experience, since the solvent change and the initial ligand choices came from the human experts.

Editorial extensions

If this is right

  • If the 72.2% top-1 retrosynthesis result is reproducible, template-free LLM retrosynthesis is competitive with dedicated graph and transformer methods on the standard 50k benchmark.
  • Yield and selectivity prediction without DFT descriptors would make data-driven optimization applicable to electronic lab notebook and literature data at lower computational cost.
  • Adding Chemma-generated yields to Bayesian optimization is reported to reach high yields in fewer experimental batches, suggesting synthetic data can substitute for some real observations.
  • If the 15-run open-space result is attributed to the model, active learning with a fine-tuned LLM can discover conditions for reaction classes absent from the training corpus, a step toward autonomous synthesis.

Reading between the lines

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

  • I read the wet-lab result as evidence for a more modest mechanism than the paper's headline: Chemma contributes candidate diversity, but the chemists supplied the initial ligand pool and the solvent switch from p-xylene to 1,4-dioxane, so the 15-run count alone does not isolate the model's causal contribution.
  • Round 0 of the open-space campaign produced yields below 15% for all nine tested ligands, suggesting the model's zero-shot extrapolation to a genuinely new reaction is weak and that the fine-tuning-in-the-loop step is the likely source of improvement.
  • A natural testable extension would be to run the same active-learning protocol with the model's suggestions replaced by random or expert-only suggestions while keeping the human intervention fixed; the difference in runs-to-target would quantify the model's added value.
  • If the shared reaction embeddings are the reason for the transfer, the same two-stage training recipe could be applied to other reaction-performance endpoints, but each new reaction family would still require its own fine-tuning and validation.
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 / 5 minor

Summary. Zhang et al. present Chemma, a fully fine-tuned LLaMA-2-7B model trained on 1.28 million reaction Q&A pairs for tasks including forward prediction, retrosynthesis, condition generation, yield and selectivity prediction, and reaction optimization. The paper claims state-of-the-art top-1 accuracy on USPTO-50k retrosynthesis (72.2%), competitive yield and selectivity predictions on HTE/ELN/literature datasets, improved Bayesian optimization when augmented with Chemma-generated yields (Chemma-BO), and an active-learning wet-lab demonstration in which an unreported Suzuki-Miyaura coupling of cyclic aminoboronates with aryl halides is optimized within 15 runs to 67% isolated yield.

Significance. If fully supported, the work would be a useful demonstration that a domain-adapted LLM can serve as a practical synthesis assistant. The paper has notable strengths: a large multi-task chemistry instruction dataset, evaluation on standard public benchmarks such as USPTO-50k and ORD-derived HTE data, a publicly accessible web service, and a real wet-lab demonstration on a previously unreported reaction type. However, the evidence as presented does not fully support the strongest claims. The wet-lab success is confounded by human expert decisions, and several benchmark comparisons lack statistical error bars or significance tests. The Chemma-BO results are in-silico on existing HTE datasets, and the evaluation protocol is not fully specified with respect to potential data leakage. The central contribution therefore needs either additional control experiments or ablations, or a careful reframing of the claims from autonomous exploration to human-AI collaboration.

major comments (4)
  1. [Exploring and optimizing open reaction spaces with Chemma; Fig. 5D and Limitations] The 15-run wet-lab demonstration does not establish that Chemma caused the observed acceleration. The text states that in round 0 "the initial combinations of starting condition variables are selected by chemists' experience" and that Chemma proposed only two of nine ligands, all of which gave yields below 15%. The improvement to 67% occurs only after the authors write "following the suggestion of chemists, we change the solvent from p-Xylene to 1,4-Dioxane." The Limitations section further concedes that "For an unreported reaction, Chemma is unable to generate effective answers without chemists' feedback (round 0 in Fig. 5D)." Because no control experiment replaces Chemma with random or expert-only ligand selection under the same human solvent rule, the claim of "autonomous experimental exploration and optimization in open reaction spaces" (abstract) is not supported. The authors should either provide such an ablation or explicitly revise the claim to describe a human-AI collaborative case study rather than autonomous model-driven exploration.
  2. [Retrosynthesis; Yield prediction; Fig. 3A and Fig. 3D] The headline claim that "Chemma surpasses the best-known results" is not supported by the reported statistics. The retrosynthesis comparison (Fig. 3A) is a single fixed test set with one model run; no standard deviations, repeated-seed results, or significance tests are reported for Chemma or the baselines. The yield-prediction comparisons (Fig. 3D) report R2 and RMSE for individual random splits, and while Extended Data Fig. 4 shows that performance varies across split strategies, no confidence intervals or statistical tests are given. For the yield claim, the manuscript does not actually name a "best-known" baseline against which Chemma is compared in the main text. Please add error bars or repeated-run statistics, and explicitly identify the state-of-the-art numbers being surpassed.
  3. [Implementation of Chemma-BO; Fig. 4E-F] The Chemma-BO evaluation is at risk of circularity because the manuscript does not specify whether the Chemma yield model used to score the full reaction space was trained on the same HTE dataset from which the observed yields are drawn. If the yield predictor has already seen many of the evaluated conditions during training, the "generated yields" are not a fair prior for a new optimization campaign. The authors should state the training split of the yield model, confirm that BO validation conditions are disjoint from its training data, and report the variance of the ten repeated runs (e.g., shaded confidence bands) in Fig. 4E-F. As written, the claim that Chemma "significantly improves" BO is not quantitatively supported.
  4. [Active learning framework of Chemma for reaction optimization; Figs. 5B-C] The active-learning workflow as described is not fully algorithmic and is therefore difficult to reproduce. The text repeatedly relies on unspecified chemist interventions, including choosing when to change base-solvent combinations, deciding between zero-shot and in-context-learning prompts, selecting when to fine-tune Chemma, and judging when a ligand suggestion should be regenerated. For example, in the C-H arylation example the authors write "we proceed to randomly select a new base-solvent combination" without specifying a rule. To support the claim of an active-learning framework, the authors should provide a precise pseudocode description of the decision rules and prompt-selection procedure, or clearly label the workflow as human-mediated exploration with Chemma as a suggestion generator.
minor comments (5)
  1. [Fig. 5C and Extended Data Fig. 8C] The text says Chemma generates AdBrettPhos in run 1 but then states "The achieved yield for this generated ligand XPhos is 4.95%"; the ligand name is inconsistent and should be corrected.
  2. [Discussion] The phrase "start-of-art performance" should be corrected to "state-of-the-art performance," and the duplicated sentence in the first paragraph of the introduction should be removed.
  3. [Fig. 3D and Performance of Chemma on open benchmark data] The text says 30% of reactions are used as an out-of-sample test set for the Suzuki-Miyaura dataset but then states that four of the 28 substrate combinations are selected as testing cases; the relationship between these two statements should be clarified.
  4. [Performance of Chemma on open benchmark data] The phrase "All reactions for evaluation are posted in Extended Data Fig. 2" should read "presented" or "shown," and the figure callouts should be checked for consistency with the referenced panels.
  5. [Supplementary prompts] Extended Data Fig. 10 contains typos such as "chemcial assistant" and "specifc chemical reaction"; the prompt templates should be carefully proofread.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: benchmark evaluations use external held-out data, and the wet-lab claim is an experimental-attribution limitation, not a self-referential derivation.

full rationale

Chemma's core predictive claims are benchmarked against external public datasets using held-out splits: retrosynthesis is trained and tested on the same USPTO-50k split as the baselines, and yield/selectivity predictions are evaluated on HTE, ELN, and literature-derived datasets from independent sources (e.g., Refs. 55, 58, 59, 62). The reported R2/RMSE values compare model outputs to observed experimental yields, so the predictions are not defined in terms of the fitted data. The Chemma-BO procedure uses Chemma-generated yields as a prior and then fits a Gaussian process to the residual between observed and Chemma-generated yields, meaning the final optimization is anchored to experiment rather than being identical to Chemma's own outputs. The open-reaction-space demonstration is the only place where the paper's framing could be challenged, but the issue is causal attribution, not circularity: the text explicitly states that the decisive solvent switch 'following the suggestion of chemists, we change the solvent from p-Xylene to 1,4-Dioxane' was made by chemists, and the Limitations section concedes that 'For an unreported reaction, Chemma is unable to generate effective answers without chemists' feedback (round 0 in Fig. 5D).' This weakens the claim that Chemma caused the 15-run acceleration and would require a control experiment, but it does not reduce a prediction to a fitted parameter or import a load-bearing result from a self-citation. No circular derivation step is exhibited in the paper.

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

This is an empirical machine learning paper, so the central claim rests on model weights and hyperparameters fitted to data, plus domain assumptions about SMILES representability, dataset quality, and the validity of the single wet-lab demonstration. No new physical or conceptual entities are introduced; Chemma and Chemma-RM are trained artifacts, not invented entities.

free parameters (4)
  • Chemma-SFT fine-tuned weights = not disclosed
    The central predictive capability is the outcome of fully fine-tuning LLaMA-2-7B on 1.28M Q&A pairs; the trained weights are the fitted parameters, hosted only behind a web service.
  • MLP regression head weights = not disclosed
    Stage-2 yield/selectivity prediction uses a five-layer MLP trained on Chemma-SFT embeddings; the weights are not released.
  • Number of fine-tuning epochs = 4
    Chemma-SFT was trained for four epochs over approximately 72 hours; this hyperparameter is chosen by hand.
  • Top-5 selections in Chemma-BO = 5
    The Bayesian optimization pipeline selects the top-5 conditions per batch based on Chemma-generated yields; this choice directly affects the reported convergence speed.
assumptions (4)
  • domain assumption SMILES sequences carry sufficient information to learn chemistry that generalizes to new reactions.
    The entire pipeline represents molecules only as SMILES strings and assumes the LLM learns valid chemical structure and reactivity from this representation; invoked throughout the Methods.
  • domain assumption Public datasets (USPTO-50k, ORD, and HTE collections) are free of synthesis errors and label noise.
    All benchmark conclusions are based on these datasets without independent curation; see Data and model availability section.
  • domain assumption The wet-lab experiments in Fig. 5D and 5E were executed correctly and the 67% isolated yield is representative.
    The wet-lab validation is a single campaign with no replication or quantitative error analysis; see 'Exploring and optimizing open reaction spaces with Chemma'.
  • domain assumption The DFT descriptors used in the BO baseline (Ahneman et al.) are a fair and sufficient representation for comparison.
    Chemma-BO and the BO baseline both rely on the same descriptors; this choice defines the reaction space in the optimization simulations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Large Language Models to Accelerate Organic Chemistry Synthesis." pith.science (2026). https://pith.science/paper/4KGPRHB2

@misc{pith2026250418340,
  author       = {Pith},
  title        = {Pith review of: Large Language Models to Accelerate Organic Chemistry Synthesis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4KGPRHB2}},
  note         = {Machine review of arXiv:2504.18340}
}
abstract

Chemical synthesis, as a foundational methodology in the creation of transformative molecules, exerts substantial influence across diverse sectors from life sciences to materials and energy. Current chemical synthesis practices emphasize laborious and costly trial-and-error workflows, underscoring the urgent need for advanced AI assistants. Nowadays, large language models (LLMs), typified by GPT-4, have been introduced as an efficient tool to facilitate scientific research. Here, we present Chemma, a fully fine-tuned LLM with 1.28 million pairs of Q&A about reactions, as an assistant to accelerate organic chemistry synthesis. Chemma surpasses the best-known results in multiple chemical tasks, e.g., single-step retrosynthesis and yield prediction, which highlights the potential of general AI for organic chemistry. Via predicting yields across the experimental reaction space, Chemma significantly improves the reaction exploration capability of Bayesian optimization. More importantly, integrated in an active learning framework, Chemma exhibits advanced potential for autonomous experimental exploration and optimization in open reaction spaces. For an unreported Suzuki-Miyaura cross-coupling reaction of cyclic aminoboronates and aryl halides for the synthesis of $\alpha$-Aryl N-heterocycles, the human-AI collaboration successfully explored suitable ligand and solvent (1,4-dioxane) within only 15 runs, achieving an isolated yield of 67%. These results reveal that, without quantum-chemical calculations, Chemma can comprehend and extract chemical insights from reaction data, in a manner akin to human experts. This work opens avenues for accelerating organic chemistry synthesis with adapted large language models.

Figures

Figures reproduced from arXiv: 2504.18340 by the authors.

Figure 1
Figure 1. Functions and applications of Chemma. (A) Illustration of using Chemma to accel￾erate organic chemistry synthesis. Scientists can interact with Chemma about four main tasks, including forward prediction, retrosynthesis, condition generation, and performance prediction (e.g., yields and selectivities). We also present an active learning framework, using Chemma as an experiment assistant, to accelerate the exploration… view at source ↗
Figure 2
Figure 2. Model architecture of Chemma and training strategies for chemical tasks. [PITH_FULL_IMAGE:figures/full_fig_p035_2.png] view at source ↗
Figure 3
Figure 3. Performance evaluation of Chemma’s capabilities for different organic synthesis [PITH_FULL_IMAGE:figures/full_fig_p036_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Value of Chemma-generated data for enhancing yield prediction and reaction op [PITH_FULL_IMAGE:figures/full_fig_p038_4.png]
Figure 5
Figure 5. Figure 5: Illustrations of using Chemma to explore reaction spaces. [PITH_FULL_IMAGE:figures/full_fig_p040_5.png]

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. Large Language Models Transform Organic Synthesis From Reaction Prediction to Automation

    cs.AI 2025-08 reject novelty 2.0 of 10

    This survey claims to be the first systematic review of LLMs for organic synthesis, but its central 'evaluation' is never actually performed.

Reference graph

Works this paper leans on

68 extracted references · 67 canonical work pages · cited by 1 Pith paper

  1. [1]

    Mendoza, Y

    A. Mendoza, Y . Ishihara, P. S. Baran, Scalable enantioselective total synthesis of taxanes, Nature Chemistry 4, 21–25 (2012)

    work page 2012

  2. [2]

    K. S. Elvira, X. C. i Solvas, R. C. Wootton, A. J. Demello, The past, present and potential for microfluidic reactor technology in chemical synthesis, Nature Chemistry 5, 905–915 (2013)

    work page 2013

  3. [3]

    Ball, Chemistry: why synthesize?, Nature 528, 327–329 (2015)

    P. Ball, Chemistry: why synthesize?, Nature 528, 327–329 (2015)

    work page 2015

  4. [4]

    S. H. Newman-Stonebraker, et al., Univariate classification of phosphine ligation state and reactivity in cross-coupling catalysis, Science 374, 301–308 (2021)

    work page 2021

  5. [5]

    Mikulak-Klucznik, et al., Computational planning of the synthesis of complex natural products, Nature 588, 83–88 (2020)

    B. Mikulak-Klucznik, et al., Computational planning of the synthesis of complex natural products, Nature 588, 83–88 (2020)

    work page 2020

  6. [6]

    K. M. Jablonka, P. Schwaller, A. Ortega-Guerrero, B. Smit, Leveraging large language models for predictive chemistry, Nature Machine Intelligence 1–9 (2024)

    work page 2024

  7. [7]

    Shen,et al., Automation and computer-assisted planning for chemical synthesis, Nature Reviews Methods Primers 1, 1–23 (2021)

    Y . Shen,et al., Automation and computer-assisted planning for chemical synthesis, Nature Reviews Methods Primers 1, 1–23 (2021)

    work page 2021

  8. [8]

    Chenthamarakshan,et al., Accelerating drug target inhibitor discovery with a deep gen- erative foundation model, Science Advances 9, eadg7865 (2023)

    V . Chenthamarakshan,et al., Accelerating drug target inhibitor discovery with a deep gen- erative foundation model, Science Advances 9, eadg7865 (2023). 24

    work page 2023

Show all 68 references
  1. [9]

    Tao, et al., Nanoparticle synthesis assisted by machine learning, Nature Reviews Mate- rials 6, 701–716 (2021)

    H. Tao, et al., Nanoparticle synthesis assisted by machine learning, Nature Reviews Mate- rials 6, 701–716 (2021)

    2021

  2. [10]

    Merchant, et al

    A. Merchant, et al. , Scaling deep learning for materials discovery, Nature 624, 80–85 (2023)

    2023

  3. [11]

    Burger, et al., A mobile robotic chemist, Nature 583, 237–241 (2020)

    B. Burger, et al., A mobile robotic chemist, Nature 583, 237–241 (2020)

    2020

  4. [12]

    N. H. Angello, et al., Closed-loop optimization of general reaction conditions for heteroaryl Suzuki-Miyaura coupling, Science 378, 399–405 (2022)

    2022

  5. [13]

    I. O. Betinol, J. Lai, S. Thakur, J. P. Reid, A data-driven workflow for assigning and pre- dicting generality in asymmetric catalysis, Journal of the American Chemical Society 145, 12870–12883 (2023)

    2023

  6. [14]

    N. I. Rinehart, et al., A machine-learning tool to predict substrate-adaptive conditions for Pd-catalyzed C–N couplings, Science 381, 965–972 (2023)

    2023

  7. [15]

    J. M. Granda, L. Donina, V . Dragone, D.-L. Long, L. Cronin, Controlling an organic synthe- sis robot with machine learning to search for new reactivity, Nature 559, 377–381 (2018)

    2018

  8. [16]

    S. H. M. Mehr, M. Craven, A. I. Leonov, G. Keenan, L. Cronin, A universal system for digitization and automatic execution of the chemical synthesis literature,Science 370, 101– 108 (2020)

    2020

  9. [17]

    Rohrbach, et al., Digitization and validation of a chemical synthesis literature database in the ChemPU, Science 377, 172–180 (2022)

    S. Rohrbach, et al., Digitization and validation of a chemical synthesis literature database in the ChemPU, Science 377, 172–180 (2022)

    2022

  10. [18]

    Sanchez-Lengeling, A

    B. Sanchez-Lengeling, A. Aspuru-Guzik, Inverse molecular design using machine learning: Generative models for matter engineering, Science 361, 360–365 (2018). 25

    2018

  11. [19]

    Wang, et al., Scientific discovery in the age of artificial intelligence, Nature 620, 47–60 (2023)

    H. Wang, et al., Scientific discovery in the age of artificial intelligence, Nature 620, 47–60 (2023)

    2023

  12. [20]

    Toniato, P

    A. Toniato, P. Schwaller, A. Cardinale, J. Geluykens, T. Laino, Unassisted noise reduction of chemical reaction datasets, Nature Machine Intelligence 3, 485–494 (2021)

    2021

  13. [21]

    Achiam, et al., GPT-4 technical report, arXiv preprint arXiv:2303.08774 (2023)

    J. Achiam, et al., GPT-4 technical report, arXiv preprint arXiv:2303.08774 (2023)

    2023 arXiv

  14. [22]

    S. A. Lehr, A. Caliskan, S. Liyanage, M. R. Banaji, Chatgpt as research scientist: Probing gpt’s capabilities as a research librarian, research ethicist, data generator, and data predictor, Proceedings of the National Academy of Sciences 121, e2404328121 (2024)

    2024

  15. [23]

    Y . Kang, J. Kim, ChatMOF: an artificial intelligence system for predicting and generating metal-organic frameworks using large language models, Nature Communications 15, 4705 (2024)

    2024

  16. [24]

    Dagdelen, et al., Structured information extraction from scientific text with large lan- guage models, Nature Communications 15, 1418 (2024)

    J. Dagdelen, et al., Structured information extraction from scientific text with large lan- guage models, Nature Communications 15, 1418 (2024)

    2024

  17. [25]

    W. Hou, Z. Ji, Assessing GPT-4 for cell type annotation in single-cell RNA-seq analysis, Nature Methods 1–4 (2024)

    2024

  18. [26]

    Zheng, et al

    Z. Zheng, et al. , A GPT-4 reticular chemist for guiding MOF discovery, Angewandte Chemie International Edition 62, e202311983 (2023)

    2023

  19. [27]

    D. A. Boiko, R. MacKnight, B. Kline, G. Gomes, Autonomous chemical research with large language models, Nature 624, 570–578 (2023)

    2023

  20. [28]

    R. B. Canty, M. Abolhasani, Reproducibility in automated chemistry laboratories using computer science abstractions, Nature Synthesis 1–13 (2024). 26

    2024

  21. [29]

    Ruan,et al., An automatic end-to-end chemical synthesis development platform powered by large language models, Nature Communications 15, 10160 (2024)

    Y . Ruan,et al., An automatic end-to-end chemical synthesis development platform powered by large language models, Nature Communications 15, 10160 (2024)

    2024

  22. [30]

    Zheng, et al., ChatGPT research group for optimizing the crystallinity of MOFs and COFs, ACS Central Science 9, 2161–2170 (2023)

    Z. Zheng, et al., ChatGPT research group for optimizing the crystallinity of MOFs and COFs, ACS Central Science 9, 2161–2170 (2023)

    2023

  23. [31]

    A. M. Bran, et al., Augmenting large language models with chemistry tools, Nature Ma- chine Intelligence 1–11 (2024)

    2024

  24. [32]

    Zheng, O

    Z. Zheng, O. Zhang, C. Borgs, J. T. Chayes, O. M. Yaghi, ChatGPT chemistry assistant for text mining and the prediction of MOF synthesis, Journal of the American Chemical Society 145, 18048–18062 (2023)

    2023

  25. [33]

    L. M. Antunes, K. T. Butler, R. Grau-Crespo, Crystal structure generation with autoregres- sive large language modeling, Nature Communications 15, 1–16 (2024)

    2024

  26. [34]

    Zheng, et al., Integrating machine learning and large language models to advance explo- ration of electrochemical reactions, Angewandte Chemie e202418074 (2024)

    Z. Zheng, et al., Integrating machine learning and large language models to advance explo- ration of electrochemical reactions, Angewandte Chemie e202418074 (2024)

    2024

  27. [35]

    M. C. Ramos, C. J. Collison, A. D. White, A review of large language models and au- tonomous agents in chemistry, Chemical Science (2025)

    2025

  28. [36]

    M. H. Segler, M. Preuss, M. P. Waller, Planning chemical syntheses with deep neural net- works and symbolic AI, Nature 555, 604–610 (2018)

    2018

  29. [37]

    C. W. Coley, et al., A robotic platform for flow synthesis of organic compounds informed by AI planning, Science eaax1566 (2019)

    2019

  30. [38]

    B. J. Shields, et al., Bayesian reaction optimization as a tool for chemical synthesis,Nature 590, 89–96 (2021). 27

    2021

  31. [39]

    T. Tang, et al., Interrogating the mechanistic features of Ni (i)-mediated aryl iodide ox- idative addition using electroanalytical and statistical modeling techniques, Journal of the American Chemical Society 145, 8689–8699 (2023)

    2023

  32. [40]

    J. Y . Wang, et al., Identifying general reaction conditions by bandit optimization, Nature 626, 1025–1033 (2024)

    2024

  33. [41]

    Raghavan, et al., Dataset design for building models of chemical reactivity,ACS Central Science 9, 2196–2204 (2023)

    P. Raghavan, et al., Dataset design for building models of chemical reactivity,ACS Central Science 9, 2196–2204 (2023)

    2023

  34. [42]

    N. C. Frey, et al., Neural scaling of deep chemical models, Nature Machine Intelligence 5, 1297–1305 (2023)

    2023

  35. [43]

    S. M. Kearnes, et al., The open reaction database, Journal of the American Chemical Soci- ety 143, 18820–18826 (2021)

    2021

  36. [44]

    C. W. Coley, L. Rogers, W. H. Green, K. F. Jensen, Computer-assisted retrosynthesis based on molecular similarity, ACS Central Science 3, 1237–1245 (2017)

    2017

  37. [45]

    D. M. Lowe, Extraction of chemical structures and reactions from the literature, Ph.D. thesis (2012)

    2012

  38. [46]

    Z. Tu, C. W. Coley, Permutation invariant graph-to-sequence model for template-free ret- rosynthesis and reaction prediction, Journal of Chemical Information and Modeling 62, 3503–3513 (2022)

    2022

  39. [47]

    Sacha, et al., Molecule edit graph attention network: modeling chemical reactions as sequences of graph edits, Journal of Chemical Information and Modeling 61, 3273–3284 (2021)

    M. Sacha, et al., Molecule edit graph attention network: modeling chemical reactions as sequences of graph edits, Journal of Chemical Information and Modeling 61, 3273–3284 (2021). 28

    2021

  40. [48]

    Seo, et al., GTA: Graph truncated attention for retrosynthesis, Proceedings of the AAAI Conference on Artificial Intelligence (2021), vol

    S.-W. Seo, et al., GTA: Graph truncated attention for retrosynthesis, Proceedings of the AAAI Conference on Artificial Intelligence (2021), vol. 35, 531–539

    2021

  41. [49]

    V . R. Somnath, C. Bunne, C. Coley, A. Krause, R. Barzilay, Learning graph models for retrosynthesis prediction, Advances in Neural Information Processing Systems 34, 9405– 9415 (2021)

    2021

  42. [50]

    Wang, et al., Retroprime: A diverse, plausible and transformer-based method for single- step retrosynthesis predictions, Chemical Engineering Journal 420, 129845 (2021)

    X. Wang, et al., Retroprime: A diverse, plausible and transformer-based method for single- step retrosynthesis predictions, Chemical Engineering Journal 420, 129845 (2021)

    2021

  43. [51]

    Wan, C.-Y

    Y . Wan, C.-Y . Hsieh, B. Liao, S. Zhang, Retroformer: Pushing the limits of end-to-end retrosynthesis transformer, International Conference on Machine Learning(2022), 22475– 22490

    2022

  44. [52]

    S. Chen, Y . Jung, Deep retrosynthetic reaction prediction using local reactivity and global attention, JACS Au 1, 1612–1620 (2021)

    2021

  45. [53]

    Yao, et al

    L. Yao, et al. , Node-aligned graph-to-graph: Elevating template-free deep learning ap- proaches in single-step retrosynthesis, JACS Au (2024)

    2024

  46. [54]

    C. W. Coley, R. Barzilay, T. S. Jaakkola, W. H. Green, K. F. Jensen, Prediction of organic reaction outcomes using machine learning, ACS Central Science 3, 434–443 (2017)

    2017

  47. [55]

    D. T. Ahneman, J. G. Estrada, S. Lin, S. D. Dreher, A. G. Doyle, Predicting reaction per- formance in C–N cross-coupling using machine learning, Science 360, 186–190 (2018)

    2018

  48. [56]

    Li, L.-C

    S.-W. Li, L.-C. Xu, C. Zhang, S.-Q. Zhang, X. Hong, Reaction performance prediction with an extrapolative and interpretable graph model based on chemical knowledge, Nature Communications 14, 3569 (2023). 29

    2023

  49. [57]

    N. J. Szymanski, et al., An autonomous laboratory for the accelerated synthesis of novel materials, Nature 624, 86–91 (2023)

    2023

  50. [58]

    Saebi, et al., On the use of real-world datasets for reaction yield prediction, Chemical Science 14, 4997–5005 (2023)

    M. Saebi, et al., On the use of real-world datasets for reaction yield prediction, Chemical Science 14, 4997–5005 (2023)

    2023

  51. [59]

    Li, X.-Q

    D.-Z. Li, X.-Q. Gong, Challenges with literature-derived data in machine learning for yield prediction: A case study on pd-catalyzed carbonylation reactions, The Journal of Physical Chemistry A 128, 10423–10430 (2024)

    2024

  52. [60]

    Li, S.-Q

    X. Li, S.-Q. Zhang, L.-C. Xu, X. Hong, Predicting regioselectivity in radical C–H func- tionalization of heterocycles through machine learning, Angewandte Chemie International Edition 59, 13253–13259 (2020)

    2020

  53. [61]

    A. F. Zahrt, et al., Prediction of higher-selectivity catalysts by computer-driven workflow and machine learning, Science 363, eaau5631 (2019)

    2019

  54. [62]

    Perera, et al

    D. Perera, et al. , A platform for automated nanomole-scale reaction screening and micromole-scale synthesis in flow, Science 359, 429–434 (2018)

    2018

  55. [63]

    Guo, et al., What can large language models do in chemistry? a comprehensive bench- mark on eight tasks, Advances in Neural Information Processing Systems36, 59662–59688 (2023)

    T. Guo, et al., What can large language models do in chemistry? a comprehensive bench- mark on eight tasks, Advances in Neural Information Processing Systems36, 59662–59688 (2023)

    2023

  56. [64]

    R. D. Taylor, M. MacCoss, A. D. Lawson, Rings in drugs: Miniperspective, Journal of Medicinal Chemistry 57, 5845–5859 (2014)

    2014

  57. [65]

    Ma, et al., A general approach to stereospecific cross-coupling reactions of nitrogen- containing stereocenters, Chem 6, 781–791 (2020)

    X. Ma, et al., A general approach to stereospecific cross-coupling reactions of nitrogen- containing stereocenters, Chem 6, 781–791 (2020). 30

    2020

  58. [66]

    X. Shu, D. Zhong, Y . Lin, X. Qin, H. Huo, Modular access to chiral α-(hetero) aryl amines via ni/photoredox-catalyzed enantioselective cross-coupling, Journal of the Ameri- can Chemical Society 144, 8797–8806 (2022)

    2022

  59. [67]

    Sarkar, S

    S. Sarkar, S. Wagulde, X. Jia, V . Gevorgyan, General and selective metal-free radicalα-c–h borylation of aliphatic amines, Chem 8, 3096–3108 (2022)

    2022

  60. [68]

    suggestion & feedback loop

    P. Ruiz-Castillo, S. L. Buchwald, Applications of Palladium-catalyzed C–N cross-coupling reactions, Chemical Reviews 116, 12564–12649 (2016). Acknowledgments We thank Prof. Kuiling Ding for valuable discussion on the design of this work, and thank SJTU AI for Science platform ...

    2016

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.