{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:47S3EUTVB7J256T6JVEL74MS34","short_pith_number":"pith:47S3EUTV","schema_version":"1.0","canonical_sha256":"e7e5b252750fd3aefa7e4d48bff192df3a81461ccc1f35cc513d5abf8ccd6deb","source":{"kind":"arxiv","id":"2502.12146","version":1},"attestation_state":"computed","paper":{"title":"Diffusion-Sharpening: Fine-tuning Diffusion Models with Denoising Trajectory Sharpening","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bin Cui, Ling Yang, Mengdi Wang, Xinchen Zhang, Ye Tian, Yunhai Tong","submitted_at":"2025-02-17T18:57:26Z","abstract_excerpt":"We propose Diffusion-Sharpening, a fine-tuning approach that enhances downstream alignment by optimizing sampling trajectories. Existing RL-based fine-tuning methods focus on single training timesteps and neglect trajectory-level alignment, while recent sampling trajectory optimization methods incur significant inference NFE costs. Diffusion-Sharpening overcomes this by using a path integral framework to select optimal trajectories during training, leveraging reward feedback, and amortizing inference costs. Our method demonstrates superior training efficiency with faster convergence, and best "},"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":"2502.12146","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-02-17T18:57:26Z","cross_cats_sorted":[],"title_canon_sha256":"83c12443a0719d9ae2dfe133f500181cfed62e10514fb50f992d4a49e17e463b","abstract_canon_sha256":"8064e4b6066760369b447b957b853669cf593124967876afaeddf63e656f1fac"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:15:39.673901Z","signature_b64":"4FleLnKkKF8YOzJJThdH+gd4jzyAPbkeH2hp6RwtuPfRke7RczZqUpZ8ij6wajoZKuGgxoCGkcLB0r054pjzBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e7e5b252750fd3aefa7e4d48bff192df3a81461ccc1f35cc513d5abf8ccd6deb","last_reissued_at":"2026-07-05T10:15:39.673391Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:15:39.673391Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Diffusion-Sharpening: Fine-tuning Diffusion Models with Denoising Trajectory Sharpening","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bin Cui, Ling Yang, Mengdi Wang, Xinchen Zhang, Ye Tian, Yunhai Tong","submitted_at":"2025-02-17T18:57:26Z","abstract_excerpt":"We propose Diffusion-Sharpening, a fine-tuning approach that enhances downstream alignment by optimizing sampling trajectories. Existing RL-based fine-tuning methods focus on single training timesteps and neglect trajectory-level alignment, while recent sampling trajectory optimization methods incur significant inference NFE costs. Diffusion-Sharpening overcomes this by using a path integral framework to select optimal trajectories during training, leveraging reward feedback, and amortizing inference costs. Our method demonstrates superior training efficiency with faster convergence, and best "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.12146","kind":"arxiv","version":1},"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/2502.12146/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":"2502.12146","created_at":"2026-07-05T10:15:39.673451+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.12146v1","created_at":"2026-07-05T10:15:39.673451+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.12146","created_at":"2026-07-05T10:15:39.673451+00:00"},{"alias_kind":"pith_short_12","alias_value":"47S3EUTVB7J2","created_at":"2026-07-05T10:15:39.673451+00:00"},{"alias_kind":"pith_short_16","alias_value":"47S3EUTVB7J256T6","created_at":"2026-07-05T10:15:39.673451+00:00"},{"alias_kind":"pith_short_8","alias_value":"47S3EUTV","created_at":"2026-07-05T10:15:39.673451+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2505.15809","citing_title":"MMaDA: Multimodal Large Diffusion Language Models","ref_index":78,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/47S3EUTVB7J256T6JVEL74MS34","json":"https://pith.science/pith/47S3EUTVB7J256T6JVEL74MS34.json","graph_json":"https://pith.science/api/pith-number/47S3EUTVB7J256T6JVEL74MS34/graph.json","events_json":"https://pith.science/api/pith-number/47S3EUTVB7J256T6JVEL74MS34/events.json","paper":"https://pith.science/paper/47S3EUTV"},"agent_actions":{"view_html":"https://pith.science/pith/47S3EUTVB7J256T6JVEL74MS34","download_json":"https://pith.science/pith/47S3EUTVB7J256T6JVEL74MS34.json","view_paper":"https://pith.science/paper/47S3EUTV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.12146&json=true","fetch_graph":"https://pith.science/api/pith-number/47S3EUTVB7J256T6JVEL74MS34/graph.json","fetch_events":"https://pith.science/api/pith-number/47S3EUTVB7J256T6JVEL74MS34/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/47S3EUTVB7J256T6JVEL74MS34/action/timestamp_anchor","attest_storage":"https://pith.science/pith/47S3EUTVB7J256T6JVEL74MS34/action/storage_attestation","attest_author":"https://pith.science/pith/47S3EUTVB7J256T6JVEL74MS34/action/author_attestation","sign_citation":"https://pith.science/pith/47S3EUTVB7J256T6JVEL74MS34/action/citation_signature","submit_replication":"https://pith.science/pith/47S3EUTVB7J256T6JVEL74MS34/action/replication_record"}},"created_at":"2026-07-05T10:15:39.673451+00:00","updated_at":"2026-07-05T10:15:39.673451+00:00"}