{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:KIKQXMHMINQSM7JRC76XONRFE5","short_pith_number":"pith:KIKQXMHM","schema_version":"1.0","canonical_sha256":"52150bb0ec4361267d3117fd77362527779a37efc50b0e0dc2af0cdbb23a7e90","source":{"kind":"arxiv","id":"2502.08808","version":2},"attestation_state":"computed","paper":{"title":"A First-order Generative Bilevel Optimization Framework for Diffusion Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"A F M Saif, Gaowen Liu, Hui Yuan, Mengdi Wang, Quan Xiao, Ramana Kompella, Tianyi Chen","submitted_at":"2025-02-12T21:44:06Z","abstract_excerpt":"Diffusion models, which iteratively denoise data samples to synthesize high-quality outputs, have achieved empirical success across domains. However, optimizing these models for downstream tasks often involves nested bilevel structures, such as tuning hyperparameters for fine-tuning tasks or noise schedules in training dynamics, where traditional bilevel methods fail due to the infinite-dimensional probability space and prohibitive sampling costs. We formalize this challenge as a generative bilevel optimization problem and address two key scenarios: (1) fine-tuning pre-trained models via an in"},"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":"2502.08808","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-02-12T21:44:06Z","cross_cats_sorted":["math.OC","stat.ML"],"title_canon_sha256":"a623098c47a0f0caa76e2439a9cb6ee47e24d6fa25a613abc89a130d5f3a9938","abstract_canon_sha256":"91bcfcdbb9b64b6621b84a50ef4ba8026ae5c17bfc5473f3c4f7dc53ece11143"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:48:47.508284Z","signature_b64":"AQO34XV2PWvN4shDyj1Qiv+EWLtkM1NjBjs1HY4wrZH2xO4hDH32bo3hnwBXfHGaaf+BJ94i52YqMmZgXThjCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"52150bb0ec4361267d3117fd77362527779a37efc50b0e0dc2af0cdbb23a7e90","last_reissued_at":"2026-07-05T11:48:47.507700Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:48:47.507700Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A First-order Generative Bilevel Optimization Framework for Diffusion Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"A F M Saif, Gaowen Liu, Hui Yuan, Mengdi Wang, Quan Xiao, Ramana Kompella, Tianyi Chen","submitted_at":"2025-02-12T21:44:06Z","abstract_excerpt":"Diffusion models, which iteratively denoise data samples to synthesize high-quality outputs, have achieved empirical success across domains. However, optimizing these models for downstream tasks often involves nested bilevel structures, such as tuning hyperparameters for fine-tuning tasks or noise schedules in training dynamics, where traditional bilevel methods fail due to the infinite-dimensional probability space and prohibitive sampling costs. We formalize this challenge as a generative bilevel optimization problem and address two key scenarios: (1) fine-tuning pre-trained models via an in"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.08808","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2502.08808/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"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":"2502.08808","created_at":"2026-07-05T11:48:47.507774+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.08808v2","created_at":"2026-07-05T11:48:47.507774+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.08808","created_at":"2026-07-05T11:48:47.507774+00:00"},{"alias_kind":"pith_short_12","alias_value":"KIKQXMHMINQS","created_at":"2026-07-05T11:48:47.507774+00:00"},{"alias_kind":"pith_short_16","alias_value":"KIKQXMHMINQSM7JR","created_at":"2026-07-05T11:48:47.507774+00:00"},{"alias_kind":"pith_short_8","alias_value":"KIKQXMHM","created_at":"2026-07-05T11:48:47.507774+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2601.16399","citing_title":"A Hessian-Free Actor-Critic Algorithm for Bi-Level Reinforcement Learning with Applications to LLM Fine-Tuning","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KIKQXMHMINQSM7JRC76XONRFE5","json":"https://pith.science/pith/KIKQXMHMINQSM7JRC76XONRFE5.json","graph_json":"https://pith.science/api/pith-number/KIKQXMHMINQSM7JRC76XONRFE5/graph.json","events_json":"https://pith.science/api/pith-number/KIKQXMHMINQSM7JRC76XONRFE5/events.json","paper":"https://pith.science/paper/KIKQXMHM"},"agent_actions":{"view_html":"https://pith.science/pith/KIKQXMHMINQSM7JRC76XONRFE5","download_json":"https://pith.science/pith/KIKQXMHMINQSM7JRC76XONRFE5.json","view_paper":"https://pith.science/paper/KIKQXMHM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.08808&json=true","fetch_graph":"https://pith.science/api/pith-number/KIKQXMHMINQSM7JRC76XONRFE5/graph.json","fetch_events":"https://pith.science/api/pith-number/KIKQXMHMINQSM7JRC76XONRFE5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KIKQXMHMINQSM7JRC76XONRFE5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KIKQXMHMINQSM7JRC76XONRFE5/action/storage_attestation","attest_author":"https://pith.science/pith/KIKQXMHMINQSM7JRC76XONRFE5/action/author_attestation","sign_citation":"https://pith.science/pith/KIKQXMHMINQSM7JRC76XONRFE5/action/citation_signature","submit_replication":"https://pith.science/pith/KIKQXMHMINQSM7JRC76XONRFE5/action/replication_record"}},"created_at":"2026-07-05T11:48:47.507774+00:00","updated_at":"2026-07-05T11:48:47.507774+00:00"}