{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:JKTMPNYOVWMSPKYUCNWQPAYNCM","short_pith_number":"pith:JKTMPNYO","schema_version":"1.0","canonical_sha256":"4aa6c7b70ead9927ab14136d07830d131c2936c606406beffe91df6f49a269d1","source":{"kind":"arxiv","id":"2503.01103","version":3},"attestation_state":"computed","paper":{"title":"Direct Discriminative Optimization: Your Likelihood-Based Visual Generative Model is Secretly a GAN Discriminator","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Guande He, Huayu Chen, Jun Zhu, Kaiwen Zheng, Ming-Yu Liu, Qinsheng Zhang, Yongxin Chen","submitted_at":"2025-03-03T02:06:22Z","abstract_excerpt":"While likelihood-based generative models, particularly diffusion and autoregressive models, have achieved remarkable fidelity in visual generation, the maximum likelihood estimation (MLE) objective, which minimizes the forward KL divergence, inherently suffers from a mode-covering tendency that limits the generation quality under limited model capacity. In this work, we propose Direct Discriminative Optimization (DDO) as a unified framework that integrates likelihood-based generative training and GAN-type discrimination to bypass this fundamental constraint by exploiting reverse KL and self-ge"},"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":"2503.01103","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-03-03T02:06:22Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"50def426818874f259cc96780511e607114aca5c8258e8e4511965a808331fa5","abstract_canon_sha256":"6a92ac2bef6d1b254b0222cd43cb87d7bc7b3157cf29e25c16b6df5fbd8bdabb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:25:08.411575Z","signature_b64":"zLXIOeLRs9fC//v5MFzuRy7g11rtWumGhRIq9TxfbxjBn+pDNScC60RhsUL58yDwbf4OKdNH8sjVn0sJz4U3Dw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4aa6c7b70ead9927ab14136d07830d131c2936c606406beffe91df6f49a269d1","last_reissued_at":"2026-07-05T11:25:08.411059Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:25:08.411059Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Direct Discriminative Optimization: Your Likelihood-Based Visual Generative Model is Secretly a GAN Discriminator","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Guande He, Huayu Chen, Jun Zhu, Kaiwen Zheng, Ming-Yu Liu, Qinsheng Zhang, Yongxin Chen","submitted_at":"2025-03-03T02:06:22Z","abstract_excerpt":"While likelihood-based generative models, particularly diffusion and autoregressive models, have achieved remarkable fidelity in visual generation, the maximum likelihood estimation (MLE) objective, which minimizes the forward KL divergence, inherently suffers from a mode-covering tendency that limits the generation quality under limited model capacity. In this work, we propose Direct Discriminative Optimization (DDO) as a unified framework that integrates likelihood-based generative training and GAN-type discrimination to bypass this fundamental constraint by exploiting reverse KL and self-ge"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.01103","kind":"arxiv","version":3},"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/2503.01103/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":"2503.01103","created_at":"2026-07-05T11:25:08.411114+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.01103v3","created_at":"2026-07-05T11:25:08.411114+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.01103","created_at":"2026-07-05T11:25:08.411114+00:00"},{"alias_kind":"pith_short_12","alias_value":"JKTMPNYOVWMS","created_at":"2026-07-05T11:25:08.411114+00:00"},{"alias_kind":"pith_short_16","alias_value":"JKTMPNYOVWMSPKYU","created_at":"2026-07-05T11:25:08.411114+00:00"},{"alias_kind":"pith_short_8","alias_value":"JKTMPNYO","created_at":"2026-07-05T11:25:08.411114+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25473","citing_title":"Causal-rCM: A Unified Teacher-Forcing and Self-Forcing Open Recipe for Autoregressive Diffusion Distillation in Streaming Video Generation and Interactive World Models","ref_index":84,"is_internal_anchor":false},{"citing_arxiv_id":"2606.20310","citing_title":"Through the PRISM: Preference Representation in Intermediate States of Video Diffusion Models","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2510.08431","citing_title":"Large Scale Diffusion Distillation via Score-Regularized Continuous-Time Consistency","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09433","citing_title":"Offline Preference Optimization for Rectified Flow with Noise-Tracked Pairs","ref_index":49,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12322","citing_title":"Self-Adversarial One Step Generation via Condition Shifting","ref_index":27,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JKTMPNYOVWMSPKYUCNWQPAYNCM","json":"https://pith.science/pith/JKTMPNYOVWMSPKYUCNWQPAYNCM.json","graph_json":"https://pith.science/api/pith-number/JKTMPNYOVWMSPKYUCNWQPAYNCM/graph.json","events_json":"https://pith.science/api/pith-number/JKTMPNYOVWMSPKYUCNWQPAYNCM/events.json","paper":"https://pith.science/paper/JKTMPNYO"},"agent_actions":{"view_html":"https://pith.science/pith/JKTMPNYOVWMSPKYUCNWQPAYNCM","download_json":"https://pith.science/pith/JKTMPNYOVWMSPKYUCNWQPAYNCM.json","view_paper":"https://pith.science/paper/JKTMPNYO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.01103&json=true","fetch_graph":"https://pith.science/api/pith-number/JKTMPNYOVWMSPKYUCNWQPAYNCM/graph.json","fetch_events":"https://pith.science/api/pith-number/JKTMPNYOVWMSPKYUCNWQPAYNCM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JKTMPNYOVWMSPKYUCNWQPAYNCM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JKTMPNYOVWMSPKYUCNWQPAYNCM/action/storage_attestation","attest_author":"https://pith.science/pith/JKTMPNYOVWMSPKYUCNWQPAYNCM/action/author_attestation","sign_citation":"https://pith.science/pith/JKTMPNYOVWMSPKYUCNWQPAYNCM/action/citation_signature","submit_replication":"https://pith.science/pith/JKTMPNYOVWMSPKYUCNWQPAYNCM/action/replication_record"}},"created_at":"2026-07-05T11:25:08.411114+00:00","updated_at":"2026-07-05T11:25:08.411114+00:00"}