{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:V7P27EQI3S6UV5HTK7U4TF4U4G","short_pith_number":"pith:V7P27EQI","schema_version":"1.0","canonical_sha256":"afdfaf9208dcbd4af4f357e9c99794e1a9c99eaa3a290bd0bcc23086d2de5bc0","source":{"kind":"arxiv","id":"2605.15557","version":1},"attestation_state":"computed","paper":{"title":"When Latent Geometry Is Not Enough: Draft-Conditioned Latent Refinement for Non-Autoregressive Text Generation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"De Shuai Zhang","submitted_at":"2026-05-15T02:53:49Z","abstract_excerpt":"Continuous diffusion and flow models are attractive for non-autoregressive text generation because they can update all positions in parallel. A major difficulty is the interface between continuous latent states and discrete tokens. This report studies a draft-conditioned latent refinement model built from a frozen BERT encoder, a parallel decoder, a denoising DraftPrior, a local FlowNet, and a learned diagonal MetricNet. Early Gaussian-start experiments showed that good latent-space metrics, such as scale matching or cosine similarity, do not guarantee good decoding. Generated latents can be c"},"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":"2605.15557","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-15T02:53:49Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"dfea514a1d2f55621f27999b3201d7221ba8eb7e32c7885b0f16fc3532539049","abstract_canon_sha256":"c3bb772efa813c7d973bd8e1b7f19f81fa52d878a9eada527148ef77b989e533"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:01:05.247140Z","signature_b64":"c/iynhlDiXUvHTdNoh/QKk7CQO0eTfpUqhv7/BVV3q5Ly6Y7HD8Ur6NYRyvLhNykHEKD7ihG8Xly1o5EXGLKCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"afdfaf9208dcbd4af4f357e9c99794e1a9c99eaa3a290bd0bcc23086d2de5bc0","last_reissued_at":"2026-05-20T00:01:05.246545Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:01:05.246545Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"When Latent Geometry Is Not Enough: Draft-Conditioned Latent Refinement for Non-Autoregressive Text Generation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"De Shuai Zhang","submitted_at":"2026-05-15T02:53:49Z","abstract_excerpt":"Continuous diffusion and flow models are attractive for non-autoregressive text generation because they can update all positions in parallel. A major difficulty is the interface between continuous latent states and discrete tokens. This report studies a draft-conditioned latent refinement model built from a frozen BERT encoder, a parallel decoder, a denoising DraftPrior, a local FlowNet, and a learned diagonal MetricNet. Early Gaussian-start experiments showed that good latent-space metrics, such as scale matching or cosine similarity, do not guarantee good decoding. Generated latents can be c"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.15557","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/2605.15557/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-19T19:34:35.263868Z","status":"skipped","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T17:41:56.090646Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"ff69b6220d4cd2ed2a4c23a14565cee85c5645c226eb8ae553e15b19e75d6818"},"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":"2605.15557","created_at":"2026-05-20T00:01:05.246640+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.15557v1","created_at":"2026-05-20T00:01:05.246640+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.15557","created_at":"2026-05-20T00:01:05.246640+00:00"},{"alias_kind":"pith_short_12","alias_value":"V7P27EQI3S6U","created_at":"2026-05-20T00:01:05.246640+00:00"},{"alias_kind":"pith_short_16","alias_value":"V7P27EQI3S6UV5HT","created_at":"2026-05-20T00:01:05.246640+00:00"},{"alias_kind":"pith_short_8","alias_value":"V7P27EQI","created_at":"2026-05-20T00:01:05.246640+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/V7P27EQI3S6UV5HTK7U4TF4U4G","json":"https://pith.science/pith/V7P27EQI3S6UV5HTK7U4TF4U4G.json","graph_json":"https://pith.science/api/pith-number/V7P27EQI3S6UV5HTK7U4TF4U4G/graph.json","events_json":"https://pith.science/api/pith-number/V7P27EQI3S6UV5HTK7U4TF4U4G/events.json","paper":"https://pith.science/paper/V7P27EQI"},"agent_actions":{"view_html":"https://pith.science/pith/V7P27EQI3S6UV5HTK7U4TF4U4G","download_json":"https://pith.science/pith/V7P27EQI3S6UV5HTK7U4TF4U4G.json","view_paper":"https://pith.science/paper/V7P27EQI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.15557&json=true","fetch_graph":"https://pith.science/api/pith-number/V7P27EQI3S6UV5HTK7U4TF4U4G/graph.json","fetch_events":"https://pith.science/api/pith-number/V7P27EQI3S6UV5HTK7U4TF4U4G/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/V7P27EQI3S6UV5HTK7U4TF4U4G/action/timestamp_anchor","attest_storage":"https://pith.science/pith/V7P27EQI3S6UV5HTK7U4TF4U4G/action/storage_attestation","attest_author":"https://pith.science/pith/V7P27EQI3S6UV5HTK7U4TF4U4G/action/author_attestation","sign_citation":"https://pith.science/pith/V7P27EQI3S6UV5HTK7U4TF4U4G/action/citation_signature","submit_replication":"https://pith.science/pith/V7P27EQI3S6UV5HTK7U4TF4U4G/action/replication_record"}},"created_at":"2026-05-20T00:01:05.246640+00:00","updated_at":"2026-05-20T00:01:05.246640+00:00"}