{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:VJCQAXKWIMKGONWVRUVT7EJEWL","short_pith_number":"pith:VJCQAXKW","schema_version":"1.0","canonical_sha256":"aa45005d5643146736d58d2b3f9124b2f5204fa5ff11abb6782b8e4f2abbcbf0","source":{"kind":"arxiv","id":"2605.19470","version":1},"attestation_state":"computed","paper":{"title":"Drifting Objectives for Refining Discrete Diffusion Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Daisuke Oba, Hiroki Furuta, Naoaki Okazaki","submitted_at":"2026-05-19T07:22:17Z","abstract_excerpt":"Discrete diffusion language models (DDLMs) generate text by iteratively denoising categorical token sequences, while recent drifting methods for continuous generators suggest that part of this sampling-time correction can instead be absorbed into training through an anti-symmetric fixed-point objective. We study how to transfer this principle to DDLMs, where the main challenge is the interface with discrete text: hard token samples are non-differentiable, and categorical predictions do not directly provide continuous samples to drift. We formulate TokenDrift, a drifting objective that lifts ca"},"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":"2605.19470","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-19T07:22:17Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"2a55fa548d8939851d20834a1ab3bfb18f0a69596fe17b8594a37af7b3ef3102","abstract_canon_sha256":"0736b8aa533240a95e670fe0f1bbaee416d59be96b789d40af59db7c67b9db62"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T01:05:47.417350Z","signature_b64":"rDRaymNiJH8V1PF81+TKKaluE10stMxCo8pqFMm2DeKQSEOYVHs+EImvu0cGpPHCiJM1U02N81eth5Xn6/QDBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"aa45005d5643146736d58d2b3f9124b2f5204fa5ff11abb6782b8e4f2abbcbf0","last_reissued_at":"2026-05-20T01:05:47.416511Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T01:05:47.416511Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Drifting Objectives for Refining Discrete Diffusion Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Daisuke Oba, Hiroki Furuta, Naoaki Okazaki","submitted_at":"2026-05-19T07:22:17Z","abstract_excerpt":"Discrete diffusion language models (DDLMs) generate text by iteratively denoising categorical token sequences, while recent drifting methods for continuous generators suggest that part of this sampling-time correction can instead be absorbed into training through an anti-symmetric fixed-point objective. We study how to transfer this principle to DDLMs, where the main challenge is the interface with discrete text: hard token samples are non-differentiable, and categorical predictions do not directly provide continuous samples to drift. We formulate TokenDrift, a drifting objective that lifts ca"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.19470","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/2605.19470/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":"2605.19470","created_at":"2026-05-20T01:05:47.416655+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.19470v1","created_at":"2026-05-20T01:05:47.416655+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.19470","created_at":"2026-05-20T01:05:47.416655+00:00"},{"alias_kind":"pith_short_12","alias_value":"VJCQAXKWIMKG","created_at":"2026-05-20T01:05:47.416655+00:00"},{"alias_kind":"pith_short_16","alias_value":"VJCQAXKWIMKGONWV","created_at":"2026-05-20T01:05:47.416655+00:00"},{"alias_kind":"pith_short_8","alias_value":"VJCQAXKW","created_at":"2026-05-20T01:05:47.416655+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/VJCQAXKWIMKGONWVRUVT7EJEWL","json":"https://pith.science/pith/VJCQAXKWIMKGONWVRUVT7EJEWL.json","graph_json":"https://pith.science/api/pith-number/VJCQAXKWIMKGONWVRUVT7EJEWL/graph.json","events_json":"https://pith.science/api/pith-number/VJCQAXKWIMKGONWVRUVT7EJEWL/events.json","paper":"https://pith.science/paper/VJCQAXKW"},"agent_actions":{"view_html":"https://pith.science/pith/VJCQAXKWIMKGONWVRUVT7EJEWL","download_json":"https://pith.science/pith/VJCQAXKWIMKGONWVRUVT7EJEWL.json","view_paper":"https://pith.science/paper/VJCQAXKW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.19470&json=true","fetch_graph":"https://pith.science/api/pith-number/VJCQAXKWIMKGONWVRUVT7EJEWL/graph.json","fetch_events":"https://pith.science/api/pith-number/VJCQAXKWIMKGONWVRUVT7EJEWL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VJCQAXKWIMKGONWVRUVT7EJEWL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VJCQAXKWIMKGONWVRUVT7EJEWL/action/storage_attestation","attest_author":"https://pith.science/pith/VJCQAXKWIMKGONWVRUVT7EJEWL/action/author_attestation","sign_citation":"https://pith.science/pith/VJCQAXKWIMKGONWVRUVT7EJEWL/action/citation_signature","submit_replication":"https://pith.science/pith/VJCQAXKWIMKGONWVRUVT7EJEWL/action/replication_record"}},"created_at":"2026-05-20T01:05:47.416655+00:00","updated_at":"2026-05-20T01:05:47.416655+00:00"}