{"paper":{"title":"Controllable Molecular Generative Foundation Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Molecular generation gains reliable control by operating on motifs rather than atoms in a diffusion process.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Meng Jiang, Tengfei Luo, Weijiang Li, Yihan Zhu, Yuhan Liu","submitted_at":"2026-05-14T19:27:01Z","abstract_excerpt":"Despite the success of foundation models in language and vision, molecular graph generation still lacks a unified framework for heterogeneous design tasks with reliable controllability. While reinforcement learning (RL) offers a natural post-training mechanism for task-specific optimization, applying it to graph generative models is hindered by the vast atom-wise action spaces and chemically invalid intermediate states. We propose \\textbf{Co}ntrollable \\textbf{Mole}cular Generative Foundation Models (CoMole), built with a unified motif-aware graph diffusion pipeline. By learning a motif-aware "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Across three heterogeneous benchmarks spanning materials and drug discovery, CoMole ranks first in controllability on all nine targets, reduces MAE by up to 48.2% relative to the strongest baselines, and maintains validity above 0.94 without rule-based correction or post-hoc filtering.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The central premise that learning a motif-aware graph space successfully transfers pretrained structural priors into controllable generation and enables RL to optimize conditional reverse policies over chemically meaningful decisions without the bottlenecks of atom-level action spaces.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CoMole uses a motif-aware graph diffusion pipeline with RL to rank first in controllability on nine targets across materials and drug benchmarks while keeping validity above 0.94 without post-processing.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Molecular generation gains reliable control by operating on motifs rather than atoms in a diffusion process.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"8cb680b513078cca04fc7cab049b86e8bd2653ad3ed2c642d1fa052be965f5b1"},"source":{"id":"2605.15354","kind":"arxiv","version":1},"verdict":{"id":"2d56abb1-3329-4c64-ae1a-7f676981fcea","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T15:33:22.937303Z","strongest_claim":"Across three heterogeneous benchmarks spanning materials and drug discovery, CoMole ranks first in controllability on all nine targets, reduces MAE by up to 48.2% relative to the strongest baselines, and maintains validity above 0.94 without rule-based correction or post-hoc filtering.","one_line_summary":"CoMole uses a motif-aware graph diffusion pipeline with RL to rank first in controllability on nine targets across materials and drug benchmarks while keeping validity above 0.94 without post-processing.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The central premise that learning a motif-aware graph space successfully transfers pretrained structural priors into controllable generation and enables RL to optimize conditional reverse policies over chemically meaningful decisions without the bottlenecks of atom-level action spaces.","pith_extraction_headline":"Molecular generation gains reliable control by operating on motifs rather than atoms in a diffusion process."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.15354/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T16:01:18.091265Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T15:41:03.329573Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T14:21:54.200138Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T13:33:22.747162Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"92c9470d412117c32bd6744bb69f63e09c79755a2a4dc73371cedf4679d7ca9f"},"references":{"count":38,"sample":[{"doi":"","year":2023,"title":"DiGress: Discrete Denoising diffusion for graph generation , author=. 2023 , eprint=","work_id":"e13fa227-cf9a-4b55-9487-57aa23b92c17","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Conditional Diffusion Based on Discrete Graph Structures for Molecular Graph Generation , author=. 2023 , eprint=","work_id":"41a68084-5d9f-404f-bad1-5ebcee6bfc27","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Graph Diffusion Transformers for Multi-Conditional Molecular Generation , author=. 2024 , eprint=","work_id":"dfaadffa-9654-4e72-989e-e40fd4416029","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2019,"title":"Junction Tree Variational Autoencoder for Molecular Graph Generation , author=. 2019 , eprint=","work_id":"6282d275-9c4c-4993-b1a8-58280b469f40","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Advances in Neural Information Processing Systems , volume=","work_id":"133d8876-939a-45a3-b693-8d3a1d71022b","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":38,"snapshot_sha256":"7397bf3417dcc746a983dc6c00eb56fc18b9586048fa26f9c2fa0f4036489080","internal_anchors":1},"formal_canon":{"evidence_count":2,"snapshot_sha256":"8c2cb68b3458c821a9eda5f5ed717b5df3443abaab3f1e42789997b89f628310"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}