{"url":"/sota/text-based-de-novo-molecule-generation-on","task":{"name":"Text-based de novo Molecule Generation","url":"/task/text-based-de-novo-molecule-generation","note":null},"dataset":{"name":"ChEBI-20","url":"/dataset/chebi-20"},"category":"Medical","categories":["Medical"],"category_note":null,"description":"Text-based de novo molecule generation involves utilizing natural language processing (NLP) techniques and chemical information to generate entirely new molecular structures. In this approach, molecular structures are typically encoded as text strings, resembling chemical formulas or SMILES (Simplified Molecular Input Line Entry System). Subsequently, by applying NLP models such as recurrent neural networks (RNNs) or Transformer models, these text strings are processed to generate novel molecular structures with desired properties.","description_from":"task","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","rank":"the archive's row order at snapshot; not re-ranked","rows_end_at":"2025-07-28","rows_withheld_as_spam":0,"metric_values":"the archive's strings, untouched"},"metrics":["BLEU","Exact Match","Frechet ChemNet Distance (FCD)","Levenshtein","MACCS FTS","Morgan FTS","RDK FTS","Text2Mol","Validity","Parameter Count"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"BLEU":"higher","Exact Match":null,"Frechet ChemNet Distance (FCD)":"lower","Levenshtein":null,"MACCS FTS":null,"Morgan FTS":null,"RDK FTS":null,"Text2Mol":null,"Validity":null,"Parameter Count":null}},"counts":{"rows":20,"rows_with_code":19,"rows_with_paper_page":20,"rows_dated":20,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"LDMol","metrics":{"BLEU":"92.6","Exact Match":"53.3","Frechet ChemNet Distance (FCD)":"0.20","Levenshtein":"6.750","MACCS FTS":"97.3","Morgan FTS":"93.1","RDK FTS":"95.0","Validity":"94.1"},"uses_additional_data":false,"paper_date":"2024-05-28","paper":"/paper/ldmol-text-conditioned-molecule-diffusion","paper_url":"https://arxiv.org/abs/2405.17829v2","paper_title":"LDMol: Text-to-Molecule Diffusion Model with Structurally Informative Latent Space","code":"https://github.com/jinhojsk515/ldmol","n_code_links":1,"syntology":{"n_ran":6,"n_unverified":3,"n_samples":9,"n_pointer_only_licence":1}},{"rank_in_archive_order":2,"model":"MolReFlect","metrics":{"BLEU":"90.3","Exact Match":"51.0","Levenshtein":"11.84","MACCS FTS":"92.9","Morgan FTS":"81.3","RDK FTS":"86.0","Validity":"97.7"},"uses_additional_data":false,"paper_date":"2024-11-22","paper":"/paper/molreflect-towards-fine-grained-in-context","paper_url":"https://arxiv.org/html/2411.14721v1","paper_title":"MolReFlect: Towards Fine-grained In-Context Alignment between Molecules and Texts","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":3,"model":"BioT5+","metrics":{"BLEU":"87.2","Exact Match":"52.2","Frechet ChemNet Distance (FCD)":"0.353","Levenshtein":"12.776","MACCS FTS":"90.7","Morgan FTS":"77.9","Parameter Count":"252000000","RDK FTS":"83.5","Text2Mol":"57.9","Validity":"100"},"uses_additional_data":false,"paper_date":"2024-02-27","paper":"/paper/biot5-towards-generalized-biological","paper_url":"https://arxiv.org/abs/2402.17810v2","paper_title":"BioT5+: Towards Generalized Biological Understanding with IUPAC Integration and Multi-task Tuning","code":"https://github.com/QizhiPei/BioT5","n_code_links":1,"syntology":null},{"rank_in_archive_order":4,"model":"BioT5","metrics":{"BLEU":"86.7","Exact Match":"41.3","Frechet ChemNet Distance (FCD)":".43","Levenshtein":"15.097","MACCS FTS":"88.6","Morgan FTS":"73.4","Parameter Count":"252000000","RDK FTS":"80.1","Text2Mol":"57.6","Validity":"100"},"uses_additional_data":false,"paper_date":"2023-10-11","paper":"/paper/biot5-enriching-cross-modal-integration-in","paper_url":"https://arxiv.org/abs/2310.07276v3","paper_title":"BioT5: Enriching Cross-modal Integration in Biology with Chemical Knowledge and Natural Language Associations","code":"https://github.com/QizhiPei/BioT5","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":5,"model":"MolReGPT (GPT-4-0413)","metrics":{"BLEU":"85.7","Exact Match":"28.0","Frechet ChemNet Distance (FCD)":"0.41","Levenshtein":"17.14","MACCS FTS":"90.3","Morgan FTS":"73.9","Parameter Count":"None","RDK FTS":"80.5","Text2Mol":"59.3","Validity":"89.9"},"uses_additional_data":false,"paper_date":"2023-06-11","paper":"/paper/empowering-molecule-discovery-for-molecule","paper_url":"https://arxiv.org/abs/2306.06615v2","paper_title":"Empowering Molecule Discovery for Molecule-Caption Translation with Large Language Models: A ChatGPT Perspective","code":"https://github.com/phenixace/molregpt","n_code_links":1,"syntology":null},{"rank_in_archive_order":6,"model":"MolT5-Large","metrics":{"BLEU":"85.4","Exact Match":"30.2","Frechet ChemNet Distance (FCD)":"1.20","Levenshtein":"16.07","MACCS FTS":"83.4","Morgan FTS":"68.4","Parameter Count":"770000000","RDK FTS":"74.6","Text2Mol":"55.4","Validity":"90.5"},"uses_additional_data":false,"paper_date":"2022-04-25","paper":"/paper/translation-between-molecules-and-natural","paper_url":"https://arxiv.org/abs/2204.11817v3","paper_title":"Translation between Molecules and Natural Language","code":"https://github.com/blender-nlp/MolT5","n_code_links":1,"syntology":{"n_ran":5,"n_unverified":0,"n_samples":5,"n_pointer_only_licence":0}},{"rank_in_archive_order":7,"model":"Text+Chem T5-augm base","metrics":{"BLEU":"85.3","Exact Match":"32.2","Frechet ChemNet Distance (FCD)":".05","Levenshtein":"16.87","MACCS FTS":"90.1","Morgan FTS":"75.7","Parameter Count":"220000000","RDK FTS":"81.6","Validity":"94.3"},"uses_additional_data":false,"paper_date":"2023-01-29","paper":"/paper/unifying-molecular-and-textual","paper_url":"https://arxiv.org/abs/2301.12586v2","paper_title":"Unifying Molecular and Textual Representations via Multi-task Language Modelling","code":"https://github.com/gt4sd/multitask_text_and_chemistry_t5","n_code_links":1,"syntology":null},{"rank_in_archive_order":8,"model":"TGM-DLM w/o corr","metrics":{"BLEU":"82.8","Exact Match":"24.2","Frechet ChemNet Distance (FCD)":"0.89","Levenshtein":"16.897","MACCS FTS":"87.4","Morgan FTS":"72.2","Parameter Count":"180000000","RDK FTS":"77.1","Text2Mol":"58.9","Validity":"78.9"},"uses_additional_data":false,"paper_date":"2024-02-20","paper":"/paper/text-guided-molecule-generation-with","paper_url":"https://arxiv.org/abs/2402.13040v1","paper_title":"Text-Guided Molecule Generation with Diffusion Language Model","code":"https://github.com/deno-v/tgm-dlm","n_code_links":1,"syntology":{"n_ran":14,"n_unverified":7,"n_samples":21,"n_pointer_only_licence":21}},{"rank_in_archive_order":9,"model":"TGM-DLM","metrics":{"BLEU":"82.6","Exact Match":"24.2","Frechet ChemNet Distance (FCD)":"0.77","Levenshtein":"17.003","MACCS FTS":"85.4","Morgan FTS":"68.8","Parameter Count":"180000000","RDK FTS":"73.9","Text2Mol":"58.1","Validity":"87.1"},"uses_additional_data":false,"paper_date":"2024-02-20","paper":"/paper/text-guided-molecule-generation-with","paper_url":"https://arxiv.org/abs/2402.13040v1","paper_title":"Text-Guided Molecule Generation with Diffusion Language Model","code":"https://github.com/deno-v/tgm-dlm","n_code_links":1,"syntology":{"n_ran":14,"n_unverified":7,"n_samples":21,"n_pointer_only_licence":21}},{"rank_in_archive_order":10,"model":"MolFM-Base","metrics":{"BLEU":"82.2","Exact Match":"21.0","Levenshtein":"19.445","MACCS FTS":"85.4","Morgan FTS":"75.8","Parameter Count":"296200000","RDK FTS":"69.7","Text2Mol":"58.3","Validity":"89.2"},"uses_additional_data":false,"paper_date":"2023-06-06","paper":"/paper/molfm-a-multimodal-molecular-foundation-model","paper_url":"https://arxiv.org/abs/2307.09484v2","paper_title":"MolFM: A Multimodal Molecular Foundation Model","code":"https://github.com/biofm/openbiomed","n_code_links":2,"syntology":null},{"rank_in_archive_order":11,"model":"Text+Chem T5-augm small","metrics":{"BLEU":"81.5","Exact Match":"19.1","Frechet ChemNet Distance (FCD)":"0.06","Levenshtein":"21.78","MACCS FTS":"86.4","Morgan FTS":"67.2","Parameter Count":"60000000","RDK FTS":"74.4","Validity":"95.1"},"uses_additional_data":false,"paper_date":"2023-01-29","paper":"/paper/unifying-molecular-and-textual","paper_url":"https://arxiv.org/abs/2301.12586v2","paper_title":"Unifying Molecular and Textual Representations via Multi-task Language Modelling","code":"https://github.com/gt4sd/multitask_text_and_chemistry_t5","n_code_links":1,"syntology":null},{"rank_in_archive_order":12,"model":"MolT5-Large-HV","metrics":{"BLEU":"81.0","Exact Match":"31.4","Frechet ChemNet Distance (FCD)":"0.44","Levenshtein":"16.758","MACCS FTS":"87.2","Morgan FTS":"72.2","Parameter Count":"770000000","RDK FTS":"78.6","Text2Mol":"59.0","Validity":"99.6"},"uses_additional_data":false,"paper_date":"2022-04-25","paper":"/paper/translation-between-molecules-and-natural","paper_url":"https://arxiv.org/abs/2204.11817v3","paper_title":"Translation between Molecules and Natural Language","code":"https://github.com/blender-nlp/MolT5","n_code_links":1,"syntology":{"n_ran":5,"n_unverified":0,"n_samples":5,"n_pointer_only_licence":0}},{"rank_in_archive_order":13,"model":"MolFM-Small","metrics":{"BLEU":"80.3","Exact Match":"16.9","Levenshtein":"20.868","MACCS FTS":"83.4","Morgan FTS":"72.1","Parameter Count":"13620000","RDK FTS":"66.2","Text2Mol":"57.3","Validity":"85.9"},"uses_additional_data":false,"paper_date":"2023-06-06","paper":"/paper/molfm-a-multimodal-molecular-foundation-model","paper_url":"https://arxiv.org/abs/2307.09484v2","paper_title":"MolFM: A Multimodal Molecular Foundation Model","code":"https://github.com/biofm/openbiomed","n_code_links":2,"syntology":null},{"rank_in_archive_order":14,"model":"MolReGPT (GPT-3.5-turbo)","metrics":{"BLEU":"79.0","Exact Match":"13.9","Frechet ChemNet Distance (FCD)":"0.57","Levenshtein":"24.91","MACCS FTS":"84.7","Morgan FTS":"62.4","RDK FTS":"70.8","Text2Mol":"57.1","Validity":"88.7"},"uses_additional_data":false,"paper_date":"2023-06-11","paper":"/paper/empowering-molecule-discovery-for-molecule","paper_url":"https://arxiv.org/abs/2306.06615v2","paper_title":"Empowering Molecule Discovery for Molecule-Caption Translation with Large Language Models: A ChatGPT Perspective","code":"https://github.com/phenixace/molregpt","n_code_links":1,"syntology":null},{"rank_in_archive_order":15,"model":"MolT5-base","metrics":{"BLEU":"76.9","Exact Match":"8.1","Frechet ChemNet Distance (FCD)":"2.18","Levenshtein":"24.458","MACCS FTS":"72.1","Morgan FTS":"52.9","Parameter Count":"220000000","RDK FTS":"58.8","Text2Mol":"49.6","Validity":"77.2"},"uses_additional_data":false,"paper_date":"2022-04-25","paper":"/paper/translation-between-molecules-and-natural","paper_url":"https://arxiv.org/abs/2204.11817v3","paper_title":"Translation between Molecules and Natural Language","code":"https://github.com/blender-nlp/MolT5","n_code_links":1,"syntology":{"n_ran":5,"n_unverified":0,"n_samples":5,"n_pointer_only_licence":0}},{"rank_in_archive_order":16,"model":"GIT-Mol-caption","metrics":{"BLEU":"75.6","Exact Match":"5.1","Levenshtein":"26.315","MACCS FTS":"73.8","Morgan FTS":"51.9","RDK FTS":"58.2","Validity":"92.8"},"uses_additional_data":false,"paper_date":"2023-08-14","paper":"/paper/git-mol-a-multi-modal-large-language-model","paper_url":"https://arxiv.org/abs/2308.06911v3","paper_title":"GIT-Mol: A Multi-modal Large Language Model for Molecular Science with Graph, Image, and Text","code":"https://github.com/ai-hpc-research-team/git-mol","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":17,"model":"MolT5-small","metrics":{"BLEU":"75.5","Exact Match":"7.9","Frechet ChemNet Distance (FCD)":"2.49","Levenshtein":"25.988","MACCS FTS":"70.3","Morgan FTS":"51.7","Parameter Count":"60000000","RDK FTS":"56.8","Text2Mol":"48.2","Validity":"72.1"},"uses_additional_data":false,"paper_date":"2022-04-25","paper":"/paper/translation-between-molecules-and-natural","paper_url":"https://arxiv.org/abs/2204.11817v3","paper_title":"Translation between Molecules and Natural Language","code":"https://github.com/blender-nlp/MolT5","n_code_links":1,"syntology":{"n_ran":5,"n_unverified":0,"n_samples":5,"n_pointer_only_licence":0}},{"rank_in_archive_order":18,"model":"Text+Chem T5 base","metrics":{"BLEU":"75","Exact Match":"21.2","Frechet ChemNet Distance (FCD)":"0.061","Levenshtein":"27.39","MACCS FTS":"87.4","Morgan FTS":"69.7","Parameter Count":"220000000","RDK FTS":"76.7","Validity":"79.2"},"uses_additional_data":false,"paper_date":"2023-01-29","paper":"/paper/unifying-molecular-and-textual","paper_url":"https://arxiv.org/abs/2301.12586v2","paper_title":"Unifying Molecular and Textual Representations via Multi-task Language Modelling","code":"https://github.com/gt4sd/multitask_text_and_chemistry_t5","n_code_links":1,"syntology":null},{"rank_in_archive_order":19,"model":"Text+Chem T5 small","metrics":{"BLEU":"73.9","Exact Match":"15.7","Frechet ChemNet Distance (FCD)":"0.066","Levenshtein":"28.54","MACCS FTS":"85.9","Morgan FTS":"66","Parameter Count":"60000000","RDK FTS":"73.6","Validity":"77.6"},"uses_additional_data":false,"paper_date":"2023-01-29","paper":"/paper/unifying-molecular-and-textual","paper_url":"https://arxiv.org/abs/2301.12586v2","paper_title":"Unifying Molecular and Textual Representations via Multi-task Language Modelling","code":"https://github.com/gt4sd/multitask_text_and_chemistry_t5","n_code_links":1,"syntology":null},{"rank_in_archive_order":20,"model":"MolXPT","metrics":{"Exact Match":"21.5","Frechet ChemNet Distance (FCD)":"0.45","MACCS FTS":"85.9","Morgan FTS":"66.7","Parameter Count":"350000000","RDK FTS":"75.7","Text2Mol":"57.8","Validity":"98.3"},"uses_additional_data":false,"paper_date":"2023-05-18","paper":"/paper/molxpt-wrapping-molecules-with-text-for","paper_url":"https://arxiv.org/abs/2305.10688v2","paper_title":"MolXPT: Wrapping Molecules with Text for Generative Pre-training","code":"https://huggingface.co/zequnl/molxpt","n_code_links":1,"syntology":null}],"since_archive":{"claim":"Results that newer papers report for their own method, placed here by Syntology. A model pointed at the cell in the paper's own table; the number was read from that cell and checked against this leaderboard's metric, dataset, split and scale; an independent check that saw this leaderboard's other rows and every other leaderboard on the same dataset accepted it. Not reviewed by the paper's authors or by the archive's editors, and not ranked against the archive rows.","extraction_file_present":true,"measurement":{"test_papers":883,"papers_with_output":881,"judged_true":108,"judged":110,"wilson95_lower":0.9361,"measured_on":"2026-09-24","frozen_commit":"0e3de0df94"},"measurement_note":"blind adjudication of accepted entries on a held-out split of archive papers, rules frozen before the test","coverage":{"sentence":"Syntology has checked 6,885 of the 9,623 papers on this site that are newer than the archive; results from the others appear after they are checked.","complete":false,"papers_newer_than_archive":9623,"papers_checked":6885,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":2737},"order":"newest first by month (arXiv date, else the arXiv-id month), then arXiv id descending","columns":[],"entries":[]},"syntology":{"read_at":"2026-09-25T09:33:49+00:00","claim":"Per row: N of M harvested code samples from that row's paper executed on a synthesized fixture; the other M-N are unverified. Not a reproduction of the row's number; not a correctness claim. n_pointer_only_licence counts samples the site points at rather than redistributes (a licence axis, independent of ran/unverified).","rows_with_graph_line":9,"rows_with_any_sample_ran":9,"distinct_papers_with_graph_line":5,"distinct_papers_with_any_sample_ran":5,"samples_over_distinct_papers":{"n_ran":27,"n_unverified":10,"n_samples":37,"n_pointer_only_licence":22,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":56,"n_unverified":17,"n_samples":73,"n_pointer_only_licence":43,"note":"row-weighted: a paper behind several rows is counted once per row; inflated relative to samples_over_distinct_papers by design, kept for readers summing the per-row syntology blocks"}}}