{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:VR75DMQ7N3BWEYDGRI2VLXTZY3","short_pith_number":"pith:VR75DMQ7","schema_version":"1.0","canonical_sha256":"ac7fd1b21f6ec36260668a3555de79c6ef602b51d9b1f2c1308a4363b250866e","source":{"kind":"arxiv","id":"2605.08659","version":2},"attestation_state":"computed","paper":{"title":"Pushing Biomolecular Utility-Diversity Frontiers with Supergroup Relative Policy Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Supergroup Relative Policy Optimization expands the utility-diversity Pareto frontier for biomolecular generators by directly rewarding set-level diversity.","cross_cats":["q-bio.BM"],"primary_cat":"cs.CE","authors_text":"Bin Feng, Hao Li, He Cao, Shenghua Gao, Xiangru Tang, Xinwu Ye, Yu Li, Zijing Liu","submitted_at":"2026-05-09T03:55:15Z","abstract_excerpt":"Biomolecular generators are often adapted with reward feedback to improve task-specific utility, but pushing utility alone can concentrate generation on a narrow family of candidates. Maintaining diversity is difficult because sample diversity is a set-level property. We introduce Supergroup Relative Policy Optimization (SGRPO), a flexible GRPO-style framework that directly constructs rewards from set-level diversity. For each condition, SGRPO samples a supergroup of candidate sets, compares their diversity under the same condition, and redistributes the group diversity reward to individual ro"},"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":true},"canonical_record":{"source":{"id":"2605.08659","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CE","submitted_at":"2026-05-09T03:55:15Z","cross_cats_sorted":["q-bio.BM"],"title_canon_sha256":"a75216692bf65b82c5279a2c9a3a6ba14a51bb279bdf004450d75f0076a70e0f","abstract_canon_sha256":"7f8fb2debb4121cb60fd8c9e1997395fc32152ee419161dea64c5c7eff8c21e7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:03:15.384754Z","signature_b64":"d0Vcl1omXwqt7CF0Q0mBn4vA2oPEsvMVT/nJ63/FV98h0H33TKaVuFIz+RKaKsoC3iYfSOtYWGPktqr411XoAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ac7fd1b21f6ec36260668a3555de79c6ef602b51d9b1f2c1308a4363b250866e","last_reissued_at":"2026-05-20T00:03:15.383856Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:03:15.383856Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Pushing Biomolecular Utility-Diversity Frontiers with Supergroup Relative Policy Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Supergroup Relative Policy Optimization expands the utility-diversity Pareto frontier for biomolecular generators by directly rewarding set-level diversity.","cross_cats":["q-bio.BM"],"primary_cat":"cs.CE","authors_text":"Bin Feng, Hao Li, He Cao, Shenghua Gao, Xiangru Tang, Xinwu Ye, Yu Li, Zijing Liu","submitted_at":"2026-05-09T03:55:15Z","abstract_excerpt":"Biomolecular generators are often adapted with reward feedback to improve task-specific utility, but pushing utility alone can concentrate generation on a narrow family of candidates. Maintaining diversity is difficult because sample diversity is a set-level property. We introduce Supergroup Relative Policy Optimization (SGRPO), a flexible GRPO-style framework that directly constructs rewards from set-level diversity. For each condition, SGRPO samples a supergroup of candidate sets, compares their diversity under the same condition, and redistributes the group diversity reward to individual ro"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Across decoding sweeps, SGRPO expands the utility-diversity Pareto frontier and achieves the best frontier-level metrics relative to pretrained generators, GRPO, and memory-assisted GRPO when applicable.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That supergroup sampling with leave-one-out diversity contributions provides a stable, unbiased signal for redistributing set-level rewards to individual rollouts without introducing artifacts in policy optimization or depending on specific diversity metrics.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SGRPO expands the utility-diversity Pareto frontier in biomolecular design by using supergroup sampling and leave-one-out diversity rewards combined with utility signals.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Supergroup Relative Policy Optimization expands the utility-diversity Pareto frontier for biomolecular generators by directly rewarding set-level diversity.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"ceb92b76a7c4a85325837a6af15e08a98e17ffb5a7a5cdb99adb097e68412a52"},"source":{"id":"2605.08659","kind":"arxiv","version":2},"verdict":{"id":"1551846e-d72c-4100-b774-317f3b8370cd","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-12T01:06:16.303088Z","strongest_claim":"Across decoding sweeps, SGRPO expands the utility-diversity Pareto frontier and achieves the best frontier-level metrics relative to pretrained generators, GRPO, and memory-assisted GRPO when applicable.","one_line_summary":"SGRPO expands the utility-diversity Pareto frontier in biomolecular design by using supergroup sampling and leave-one-out diversity rewards combined with utility signals.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That supergroup sampling with leave-one-out diversity contributions provides a stable, unbiased signal for redistributing set-level rewards to individual rollouts without introducing artifacts in policy optimization or depending on specific diversity metrics.","pith_extraction_headline":"Supergroup Relative Policy Optimization expands the utility-diversity Pareto frontier for biomolecular generators by directly rewarding set-level diversity."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.08659/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-19T22:36:35.227881Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T14:31:17.953922Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T10:52:18.332137Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"5b843a286c965236fd2931f5009534692dfe09cf8d1792761f3fcd76db037c3c"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"3d287ef768fa55a71c505bf91476c8782ffa57efd01fdb1d0307e2bbb2e94773"},"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":"2605.08659","created_at":"2026-05-20T00:03:15.384008+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.08659v2","created_at":"2026-05-20T00:03:15.384008+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.08659","created_at":"2026-05-20T00:03:15.384008+00:00"},{"alias_kind":"pith_short_12","alias_value":"VR75DMQ7N3BW","created_at":"2026-05-20T00:03:15.384008+00:00"},{"alias_kind":"pith_short_16","alias_value":"VR75DMQ7N3BWEYDG","created_at":"2026-05-20T00:03:15.384008+00:00"},{"alias_kind":"pith_short_8","alias_value":"VR75DMQ7","created_at":"2026-05-20T00:03:15.384008+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VR75DMQ7N3BWEYDGRI2VLXTZY3","json":"https://pith.science/pith/VR75DMQ7N3BWEYDGRI2VLXTZY3.json","graph_json":"https://pith.science/api/pith-number/VR75DMQ7N3BWEYDGRI2VLXTZY3/graph.json","events_json":"https://pith.science/api/pith-number/VR75DMQ7N3BWEYDGRI2VLXTZY3/events.json","paper":"https://pith.science/paper/VR75DMQ7"},"agent_actions":{"view_html":"https://pith.science/pith/VR75DMQ7N3BWEYDGRI2VLXTZY3","download_json":"https://pith.science/pith/VR75DMQ7N3BWEYDGRI2VLXTZY3.json","view_paper":"https://pith.science/paper/VR75DMQ7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.08659&json=true","fetch_graph":"https://pith.science/api/pith-number/VR75DMQ7N3BWEYDGRI2VLXTZY3/graph.json","fetch_events":"https://pith.science/api/pith-number/VR75DMQ7N3BWEYDGRI2VLXTZY3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VR75DMQ7N3BWEYDGRI2VLXTZY3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VR75DMQ7N3BWEYDGRI2VLXTZY3/action/storage_attestation","attest_author":"https://pith.science/pith/VR75DMQ7N3BWEYDGRI2VLXTZY3/action/author_attestation","sign_citation":"https://pith.science/pith/VR75DMQ7N3BWEYDGRI2VLXTZY3/action/citation_signature","submit_replication":"https://pith.science/pith/VR75DMQ7N3BWEYDGRI2VLXTZY3/action/replication_record"}},"created_at":"2026-05-20T00:03:15.384008+00:00","updated_at":"2026-05-20T00:03:15.384008+00:00"}