{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:U5ZII53PUDBI3ZPHQWNR4376YF","short_pith_number":"pith:U5ZII53P","schema_version":"1.0","canonical_sha256":"a77284776fa0c28de5e7859b1e6ffec142b4c702f1634c29a92b4b09d8da9e09","source":{"kind":"arxiv","id":"2604.23061","version":2},"attestation_state":"computed","paper":{"title":"C-MORAL: Controllable Multi-Objective Molecular Optimization with Reinforcement Alignment for LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Reinforcement learning post-training with group-based optimization and non-linear rewards aligns LLMs to optimize molecules across multiple competing properties.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Morteza Ziyadi, Rui Gao, Swastik Roy, Xiang 'Anthony' Chen, Youngseung Jeon","submitted_at":"2026-04-24T23:11:44Z","abstract_excerpt":"Large language models (LLMs) show promise for molecular optimization, but aligning them with selective and competing drug-design constraints remains challenging. We propose C-Moral, a reinforcement learning post-training framework for controllable multi-objective molecular optimization. C-Moral combines group-based relative optimization, property score alignment for heterogeneous objectives, and bottleneck-sensitive non-linear reward aggregation to improve stability across competing molecular properties. Experiments on C-MuMOInstruct and S$^2$-Bench MolOpt show that C-Moral achieves the best p"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2604.23061","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-04-24T23:11:44Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"26fe96969ff62fbc8062cc7d547f2bc67ec548d8cb116d88c4db89722698a604","abstract_canon_sha256":"f672d682a77307538cd790975e1ceeb9b292fdfdf0b6e93ac9090019acd7d1a0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:04:08.587206Z","signature_b64":"S781j/cMVMUtBWhDpspbmmAmzEZWO4Zhk6LNt9x19ahWh4vdQRFa3WohvOqH7+8TX/idC70Vd11K2fEWniDjAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a77284776fa0c28de5e7859b1e6ffec142b4c702f1634c29a92b4b09d8da9e09","last_reissued_at":"2026-05-28T01:04:08.586605Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:04:08.586605Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"C-MORAL: Controllable Multi-Objective Molecular Optimization with Reinforcement Alignment for LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Reinforcement learning post-training with group-based optimization and non-linear rewards aligns LLMs to optimize molecules across multiple competing properties.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Morteza Ziyadi, Rui Gao, Swastik Roy, Xiang 'Anthony' Chen, Youngseung Jeon","submitted_at":"2026-04-24T23:11:44Z","abstract_excerpt":"Large language models (LLMs) show promise for molecular optimization, but aligning them with selective and competing drug-design constraints remains challenging. We propose C-Moral, a reinforcement learning post-training framework for controllable multi-objective molecular optimization. C-Moral combines group-based relative optimization, property score alignment for heterogeneous objectives, and bottleneck-sensitive non-linear reward aggregation to improve stability across competing molecular properties. Experiments on C-MuMOInstruct and S$^2$-Bench MolOpt show that C-Moral achieves the best p"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experiments on the C-MuMOInstruct benchmark show that C-Moral consistently outperforms state-of-the-art models across both in-domain and out-of-domain settings, achieving the best Success Optimized Rate (SOR) of 48.9% on IND tasks and 39.5% on OOD tasks, while largely preserving scaffold similarity.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the reported performance gains on the C-MuMOInstruct benchmark are attributable to the proposed components (group-based relative optimization, property score alignment, and continuous non-linear reward aggregation) rather than implementation details or benchmark-specific artifacts, and that these gains generalize to practical drug design.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"C-MORAL applies reinforcement learning post-training with group-based optimization and non-linear reward aggregation to align LLMs for controllable multi-objective molecular optimization, achieving 48.9% SOR on in-domain and 39.5% on out-of-domain benchmark tasks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Reinforcement learning post-training with group-based optimization and non-linear rewards aligns LLMs to optimize molecules across multiple competing properties.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"657467683ba83042f4dcf01cc6cc105b90e1b61a083cfc5c3bc2a97de78d7259"},"source":{"id":"2604.23061","kind":"arxiv","version":2},"verdict":{"id":"d9398012-17c5-4c2a-b2bf-48766a493a30","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T11:55:55.843945Z","strongest_claim":"Experiments on the C-MuMOInstruct benchmark show that C-Moral consistently outperforms state-of-the-art models across both in-domain and out-of-domain settings, achieving the best Success Optimized Rate (SOR) of 48.9% on IND tasks and 39.5% on OOD tasks, while largely preserving scaffold similarity.","one_line_summary":"C-MORAL applies reinforcement learning post-training with group-based optimization and non-linear reward aggregation to align LLMs for controllable multi-objective molecular optimization, achieving 48.9% SOR on in-domain and 39.5% on out-of-domain benchmark tasks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the reported performance gains on the C-MuMOInstruct benchmark are attributable to the proposed components (group-based relative optimization, property score alignment, and continuous non-linear reward aggregation) rather than implementation details or benchmark-specific artifacts, and that these gains generalize to practical drug design.","pith_extraction_headline":"Reinforcement learning post-training with group-based optimization and non-linear rewards aligns LLMs to optimize molecules across multiple competing properties."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.23061/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-21T09:40:14.227532Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T23:32:21.434242Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"25f72fb6a56f92bbb4ebade731bfc90899b4b01f3d120cf23222560e699cec42"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2604.23061","created_at":"2026-05-28T01:04:08.586672+00:00"},{"alias_kind":"arxiv_version","alias_value":"2604.23061v2","created_at":"2026-05-28T01:04:08.586672+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.23061","created_at":"2026-05-28T01:04:08.586672+00:00"},{"alias_kind":"pith_short_12","alias_value":"U5ZII53PUDBI","created_at":"2026-05-28T01:04:08.586672+00:00"},{"alias_kind":"pith_short_16","alias_value":"U5ZII53PUDBI3ZPH","created_at":"2026-05-28T01:04:08.586672+00:00"},{"alias_kind":"pith_short_8","alias_value":"U5ZII53P","created_at":"2026-05-28T01:04:08.586672+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/U5ZII53PUDBI3ZPHQWNR4376YF","json":"https://pith.science/pith/U5ZII53PUDBI3ZPHQWNR4376YF.json","graph_json":"https://pith.science/api/pith-number/U5ZII53PUDBI3ZPHQWNR4376YF/graph.json","events_json":"https://pith.science/api/pith-number/U5ZII53PUDBI3ZPHQWNR4376YF/events.json","paper":"https://pith.science/paper/U5ZII53P"},"agent_actions":{"view_html":"https://pith.science/pith/U5ZII53PUDBI3ZPHQWNR4376YF","download_json":"https://pith.science/pith/U5ZII53PUDBI3ZPHQWNR4376YF.json","view_paper":"https://pith.science/paper/U5ZII53P","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2604.23061&json=true","fetch_graph":"https://pith.science/api/pith-number/U5ZII53PUDBI3ZPHQWNR4376YF/graph.json","fetch_events":"https://pith.science/api/pith-number/U5ZII53PUDBI3ZPHQWNR4376YF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/U5ZII53PUDBI3ZPHQWNR4376YF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/U5ZII53PUDBI3ZPHQWNR4376YF/action/storage_attestation","attest_author":"https://pith.science/pith/U5ZII53PUDBI3ZPHQWNR4376YF/action/author_attestation","sign_citation":"https://pith.science/pith/U5ZII53PUDBI3ZPHQWNR4376YF/action/citation_signature","submit_replication":"https://pith.science/pith/U5ZII53PUDBI3ZPHQWNR4376YF/action/replication_record"}},"created_at":"2026-05-28T01:04:08.586672+00:00","updated_at":"2026-05-28T01:04:08.586672+00:00"}