{"paper":{"title":"CoRe-Gen: Robust Spectrum-to-Structure Generation under Imperfect Fingerprint Conditions","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"CoRe-Gen generates molecular structures from mass spectra by training decoders on frequency-aware corrupted fingerprints to match real prediction noise.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Chixiang Lu, Haibo Jiang, Hengyu Zhang, Jing Hao, Lifei Wang, Tianbo Liu, Xiaojuan Qi","submitted_at":"2026-05-13T04:22:55Z","abstract_excerpt":"Molecular structure elucidation from tandem mass spectra (MS/MS) remains challenging, particularly for de novo generation beyond database coverage. A common approach decomposes the task into spectrum-to-fingerprint prediction followed by fingerprint-to-structure decoding, enabling the use of large-scale molecular corpora. However, at deployment, the decoder relies on predicted rather than oracle fingerprints, introducing structured errors that propagate into generation. This results in a fundamental condition mismatch, where models trained on clean inputs must operate under noisy, biased predi"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"CoRe-Gen establishes a new state of the art on NPLIB1, achieving 19.54% Top-1 and 29.92% Top-10 exact-match accuracy, while remaining competitive on the more challenging MassSpecGym benchmark.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that frequency-aware fingerprint corruption during training accurately reproduces the structured errors that arise from real spectrum-to-fingerprint predictors, and that the reported gains are not driven by benchmark-specific tuning or unstated data splits.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CoRe-Gen reaches new state-of-the-art exact-match accuracy on the NPLIB1 benchmark for de novo molecular structure generation from mass spectra by using synthetic pretraining, frequency-aware corruption, and structure-aware decoding to close the gap between clean training data and noisy deployment.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"CoRe-Gen generates molecular structures from mass spectra by training decoders on frequency-aware corrupted fingerprints to match real prediction noise.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"f533e9393ca050c05eb45abf1c92ebb10f9ecb83e8e6be3a96ec9720bc9075ef"},"source":{"id":"2605.12980","kind":"arxiv","version":1},"verdict":{"id":"9dc327d6-ab57-424d-96fa-3c868e5b878a","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T20:11:41.233523Z","strongest_claim":"CoRe-Gen establishes a new state of the art on NPLIB1, achieving 19.54% Top-1 and 29.92% Top-10 exact-match accuracy, while remaining competitive on the more challenging MassSpecGym benchmark.","one_line_summary":"CoRe-Gen reaches new state-of-the-art exact-match accuracy on the NPLIB1 benchmark for de novo molecular structure generation from mass spectra by using synthetic pretraining, frequency-aware corruption, and structure-aware decoding to close the gap between clean training data and noisy deployment.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that frequency-aware fingerprint corruption during training accurately reproduces the structured errors that arise from real spectrum-to-fingerprint predictors, and that the reported gains are not driven by benchmark-specific tuning or unstated data splits.","pith_extraction_headline":"CoRe-Gen generates molecular structures from mass spectra by training decoders on frequency-aware corrupted fingerprints to match real prediction noise."},"references":{"count":43,"sample":[{"doi":"","year":2024,"title":"Martin Alberts, Oliver Schilter, Fabio Zipoli, et al. 2024. Unraveling molecular structure: A multimodal spectroscopic dataset for chemistry.Advances in Neural Information Processing Systems, 37:12578","work_id":"fe97d45a-4b09-4e77-aabb-79b7a83fe604","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2014,"title":"Felix Allen, Allison Pon, Michael Wilson, et al. 2014. CFM-ID: a web server for annotation, spectrum prediction and metabolite identification from tandem mass spectra.Nucleic Acids Research, 42(W1):W9","work_id":"933be7e5-dd9c-4d86-9a3c-f142fde6c936","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"Liu Cao, Mustafa Guler, Azat Tagirdzhanov, et al. 2021. MolDiscovery: learning mass spec- trometry fragmentation of small molecules.Nature Communications, 12(1):3718","work_id":"3fdcde7c-a9bd-42a8-8131-70e5fcc1e88c","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Thomas Butler, Abraham Frandsen, Rose Lightheart, et al. 2023. MS2Mol: A transformer model for illuminating dark chemical space from mass spectra","work_id":"020ae707-97de-4930-85b5-3d29f0f2d8d9","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Bushuiev, R., Bushuiev, A., de Jonge, N","work_id":"8c64a288-f39e-4cfd-b6ec-2a5af3656597","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":43,"snapshot_sha256":"481b3b23aea0940704b34d975ad0bc2cfa770ea31e3cee818efe4c8bccb99aee","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}