{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:XH4SJKOIBKWWQVYAX6DXHVR7RL","short_pith_number":"pith:XH4SJKOI","schema_version":"1.0","canonical_sha256":"b9f924a9c80aad685700bf8773d63f8aed66f7aaa8e94b2163937f2a2e6940d8","source":{"kind":"arxiv","id":"2605.18817","version":1},"attestation_state":"computed","paper":{"title":"Multi-Token Residual Prediction","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Ang Li, Bowen Peng, Haoqi Zhang, Qian Wang, Rahul Chalamala, Yucheng Lu, Yufeng Xu, Zeshen Zhang, Zishuo Bao","submitted_at":"2026-05-12T11:40:39Z","abstract_excerpt":"Diffusion Language Models (DLMs) generate text by iteratively denoising masked token sequences, offering a tradeoff between parallelism and quality compared to autoregressive models. In current practice, the number of tokens decoded per step is controlled by a confidence threshold, and quality degrades monotonically as more tokens are denoised per step. We introduce Multi-token Residual Prediction (MRP), a lightweight module that enables dependency-aware multi-token denoising within a single backbone forward pass. MRP exploits a key property of the denoising process: the logit distributions at"},"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":"2605.18817","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-12T11:40:39Z","cross_cats_sorted":[],"title_canon_sha256":"fe54869ad43d9f3426c7ba13e880bccef6890cfb2494f98eb91af03134b542d0","abstract_canon_sha256":"19ddf6df26e4bb421ad2d7cf37a6b167dedd3f4c317e8bc7f521d11cf3d60948"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:06:24.096682Z","signature_b64":"teqYWRnNOFjM1IlUBDU3I3BlgBJgZt+9ABZVpe/XF5IWJYGK6cHkYT4TXoJY/nLjHKkeqbKSGiNHOXNud4ukBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b9f924a9c80aad685700bf8773d63f8aed66f7aaa8e94b2163937f2a2e6940d8","last_reissued_at":"2026-05-20T00:06:24.095841Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:06:24.095841Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Multi-Token Residual Prediction","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Ang Li, Bowen Peng, Haoqi Zhang, Qian Wang, Rahul Chalamala, Yucheng Lu, Yufeng Xu, Zeshen Zhang, Zishuo Bao","submitted_at":"2026-05-12T11:40:39Z","abstract_excerpt":"Diffusion Language Models (DLMs) generate text by iteratively denoising masked token sequences, offering a tradeoff between parallelism and quality compared to autoregressive models. In current practice, the number of tokens decoded per step is controlled by a confidence threshold, and quality degrades monotonically as more tokens are denoised per step. We introduce Multi-token Residual Prediction (MRP), a lightweight module that enables dependency-aware multi-token denoising within a single backbone forward pass. MRP exploits a key property of the denoising process: the logit distributions at"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.18817","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/2605.18817/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":"2605.18817","created_at":"2026-05-20T00:06:24.095997+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.18817v1","created_at":"2026-05-20T00:06:24.095997+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.18817","created_at":"2026-05-20T00:06:24.095997+00:00"},{"alias_kind":"pith_short_12","alias_value":"XH4SJKOIBKWW","created_at":"2026-05-20T00:06:24.095997+00:00"},{"alias_kind":"pith_short_16","alias_value":"XH4SJKOIBKWWQVYA","created_at":"2026-05-20T00:06:24.095997+00:00"},{"alias_kind":"pith_short_8","alias_value":"XH4SJKOI","created_at":"2026-05-20T00:06:24.095997+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/XH4SJKOIBKWWQVYAX6DXHVR7RL","json":"https://pith.science/pith/XH4SJKOIBKWWQVYAX6DXHVR7RL.json","graph_json":"https://pith.science/api/pith-number/XH4SJKOIBKWWQVYAX6DXHVR7RL/graph.json","events_json":"https://pith.science/api/pith-number/XH4SJKOIBKWWQVYAX6DXHVR7RL/events.json","paper":"https://pith.science/paper/XH4SJKOI"},"agent_actions":{"view_html":"https://pith.science/pith/XH4SJKOIBKWWQVYAX6DXHVR7RL","download_json":"https://pith.science/pith/XH4SJKOIBKWWQVYAX6DXHVR7RL.json","view_paper":"https://pith.science/paper/XH4SJKOI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.18817&json=true","fetch_graph":"https://pith.science/api/pith-number/XH4SJKOIBKWWQVYAX6DXHVR7RL/graph.json","fetch_events":"https://pith.science/api/pith-number/XH4SJKOIBKWWQVYAX6DXHVR7RL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XH4SJKOIBKWWQVYAX6DXHVR7RL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XH4SJKOIBKWWQVYAX6DXHVR7RL/action/storage_attestation","attest_author":"https://pith.science/pith/XH4SJKOIBKWWQVYAX6DXHVR7RL/action/author_attestation","sign_citation":"https://pith.science/pith/XH4SJKOIBKWWQVYAX6DXHVR7RL/action/citation_signature","submit_replication":"https://pith.science/pith/XH4SJKOIBKWWQVYAX6DXHVR7RL/action/replication_record"}},"created_at":"2026-05-20T00:06:24.095997+00:00","updated_at":"2026-05-20T00:06:24.095997+00:00"}