{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:35Z74LHNYCZSQI3MIXFX5VMKZM","short_pith_number":"pith:35Z74LHN","schema_version":"1.0","canonical_sha256":"df73fe2cedc0b328236c45cb7ed58acb189c39e6b8a3c5b87b042b40a832ff0d","source":{"kind":"arxiv","id":"2608.03929","version":1},"attestation_state":"computed","paper":{"title":"Latent Reward Registers for Diffusion Preference Alignment","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Peiqin Sun, Yuanshen Guan, Zhiwei Xiong, Zipeng Feng","submitted_at":"2026-08-04T17:00:52Z","abstract_excerpt":"Aligning diffusion models with human preferences usually relies on a sparse terminal reward evaluated on the final generated samples, presenting a severe temporal credit-assignment challenge across the multi-step denoising process. We propose Latent Reward Registers, a mechanism that estimates terminal preference directly from intermediate noisy latents by prepending learnable, position-free register tokens to the input sequence of a frozen Diffusion Transformer (DiT). This independent readout mechanism extracts latent reward evidence without altering the generator's hidden states or velocity "},"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":"2608.03929","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-08-04T17:00:52Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"d7f9632bc93d0513b2e1ad97034746182395103b0654e0c56de0b784057ebc59","abstract_canon_sha256":"adca239f1d278bd720f0d16849a52f81840178a168dcb3796ff128ba5c3d6625"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-05T01:37:31.896497Z","signature_b64":"dYz3xQq+JcrtQyAGrzQ7oaPDWebztFyT9KSseYReYhu/ZSRP7EvDKM4OOat7yzBf6bRJ9QMKhwVU2zU7PA7WCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"df73fe2cedc0b328236c45cb7ed58acb189c39e6b8a3c5b87b042b40a832ff0d","last_reissued_at":"2026-08-05T01:37:31.895113Z","signature_status":"signed_v1","first_computed_at":"2026-08-05T01:37:31.895113Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Latent Reward Registers for Diffusion Preference Alignment","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Peiqin Sun, Yuanshen Guan, Zhiwei Xiong, Zipeng Feng","submitted_at":"2026-08-04T17:00:52Z","abstract_excerpt":"Aligning diffusion models with human preferences usually relies on a sparse terminal reward evaluated on the final generated samples, presenting a severe temporal credit-assignment challenge across the multi-step denoising process. We propose Latent Reward Registers, a mechanism that estimates terminal preference directly from intermediate noisy latents by prepending learnable, position-free register tokens to the input sequence of a frozen Diffusion Transformer (DiT). This independent readout mechanism extracts latent reward evidence without altering the generator's hidden states or velocity "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.03929","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/2608.03929/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":"2608.03929","created_at":"2026-08-05T01:37:31.895594+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.03929v1","created_at":"2026-08-05T01:37:31.895594+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.03929","created_at":"2026-08-05T01:37:31.895594+00:00"},{"alias_kind":"pith_short_12","alias_value":"35Z74LHNYCZS","created_at":"2026-08-05T01:37:31.895594+00:00"},{"alias_kind":"pith_short_16","alias_value":"35Z74LHNYCZSQI3M","created_at":"2026-08-05T01:37:31.895594+00:00"},{"alias_kind":"pith_short_8","alias_value":"35Z74LHN","created_at":"2026-08-05T01:37:31.895594+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/35Z74LHNYCZSQI3MIXFX5VMKZM","json":"https://pith.science/pith/35Z74LHNYCZSQI3MIXFX5VMKZM.json","graph_json":"https://pith.science/api/pith-number/35Z74LHNYCZSQI3MIXFX5VMKZM/graph.json","events_json":"https://pith.science/api/pith-number/35Z74LHNYCZSQI3MIXFX5VMKZM/events.json","paper":"https://pith.science/paper/35Z74LHN"},"agent_actions":{"view_html":"https://pith.science/pith/35Z74LHNYCZSQI3MIXFX5VMKZM","download_json":"https://pith.science/pith/35Z74LHNYCZSQI3MIXFX5VMKZM.json","view_paper":"https://pith.science/paper/35Z74LHN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.03929&json=true","fetch_graph":"https://pith.science/api/pith-number/35Z74LHNYCZSQI3MIXFX5VMKZM/graph.json","fetch_events":"https://pith.science/api/pith-number/35Z74LHNYCZSQI3MIXFX5VMKZM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/35Z74LHNYCZSQI3MIXFX5VMKZM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/35Z74LHNYCZSQI3MIXFX5VMKZM/action/storage_attestation","attest_author":"https://pith.science/pith/35Z74LHNYCZSQI3MIXFX5VMKZM/action/author_attestation","sign_citation":"https://pith.science/pith/35Z74LHNYCZSQI3MIXFX5VMKZM/action/citation_signature","submit_replication":"https://pith.science/pith/35Z74LHNYCZSQI3MIXFX5VMKZM/action/replication_record"}},"created_at":"2026-08-05T01:37:31.895594+00:00","updated_at":"2026-08-05T01:37:31.895594+00:00"}