{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:JMSRQJSLUDDSPL2LTQ2TUBC53Q","short_pith_number":"pith:JMSRQJSL","schema_version":"1.0","canonical_sha256":"4b2518264ba0c727af4b9c353a045ddc0a6bc23dd97f5532b018a95aece41db0","source":{"kind":"arxiv","id":"2508.14148","version":2},"attestation_state":"computed","paper":{"title":"DPad: Efficient Diffusion Language Models with Suffix Dropout","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Chiyue Wei, Cong Guo, Hai \"Helen\" Li, Jianyi Zhang, Sitao Huang, Xinhua Chen, Yintao He, Yiran Chen","submitted_at":"2025-08-19T16:56:51Z","abstract_excerpt":"Diffusion-based Large Language Models (dLLMs) parallelize text generation by framing decoding as a denoising process, but suffer from high computational overhead since they predict all future suffix tokens at each step while retaining only a small fraction. We propose Diffusion Scratchpad (DPad), a training-free method that restricts attention to a small set of nearby suffix tokens, preserving fidelity while eliminating redundancy. DPad integrates two strategies: (i) a sliding window, which maintains a fixed-length suffix window, and (ii) distance-decay dropout, which deterministically removes"},"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":"2508.14148","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-08-19T16:56:51Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"9aaf646c15d6eadefc3ec1d0123c14abf8989ed4b01bec2e72d5babd920bb0b8","abstract_canon_sha256":"f1854c0810e325181b3b111ab75300ba7d976d7e456e82108576ed51b9da3ec9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:57:59.201665Z","signature_b64":"LH4LkT6eiaxFH/D0yUoxjgtdbD/hhUc9fZ3z6w6CV47NKkIhmGztvsilVvVNq/iKi8cDtVTa+cZ2jFx3jlkZCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4b2518264ba0c727af4b9c353a045ddc0a6bc23dd97f5532b018a95aece41db0","last_reissued_at":"2026-07-05T11:57:59.201190Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:57:59.201190Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DPad: Efficient Diffusion Language Models with Suffix Dropout","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Chiyue Wei, Cong Guo, Hai \"Helen\" Li, Jianyi Zhang, Sitao Huang, Xinhua Chen, Yintao He, Yiran Chen","submitted_at":"2025-08-19T16:56:51Z","abstract_excerpt":"Diffusion-based Large Language Models (dLLMs) parallelize text generation by framing decoding as a denoising process, but suffer from high computational overhead since they predict all future suffix tokens at each step while retaining only a small fraction. We propose Diffusion Scratchpad (DPad), a training-free method that restricts attention to a small set of nearby suffix tokens, preserving fidelity while eliminating redundancy. DPad integrates two strategies: (i) a sliding window, which maintains a fixed-length suffix window, and (ii) distance-decay dropout, which deterministically removes"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.14148","kind":"arxiv","version":2},"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/2508.14148/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":"2508.14148","created_at":"2026-07-05T11:57:59.201247+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.14148v2","created_at":"2026-07-05T11:57:59.201247+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.14148","created_at":"2026-07-05T11:57:59.201247+00:00"},{"alias_kind":"pith_short_12","alias_value":"JMSRQJSLUDDS","created_at":"2026-07-05T11:57:59.201247+00:00"},{"alias_kind":"pith_short_16","alias_value":"JMSRQJSLUDDSPL2L","created_at":"2026-07-05T11:57:59.201247+00:00"},{"alias_kind":"pith_short_8","alias_value":"JMSRQJSL","created_at":"2026-07-05T11:57:59.201247+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2602.08404","citing_title":"TEAM: Temporal-Spatial Consistency Guided Expert Activation for MoE Diffusion Language Model Acceleration","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14530","citing_title":"Mitigating Mask Prior Drift and Positional Attention Collapse in Large Diffusion Vision-Language Models","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18165","citing_title":"Elastic-dLLM: Position Preserving Context Compression and Augmentation of Diffusion LLMs","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08302","citing_title":"DMax: Aggressive Parallel Decoding for dLLMs","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2508.19982","citing_title":"Diffusion Language Models Know the Answer Before Decoding","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2509.20624","citing_title":"FS-DFM: Fast and Accurate Long Text Generation with Few-Step Diffusion Language Models","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2510.06133","citing_title":"CreditDecoding: Accelerating Parallel Decoding in Diffusion Large Language Models with Trace Credit","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14530","citing_title":"Mitigating Mask Prior Drift and Positional Attention Collapse in Large Diffusion Vision-Language Models","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2512.15745","citing_title":"LLaDA2.0: Scaling Up Diffusion Language Models to 100B","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08302","citing_title":"DMax: Aggressive Parallel Decoding for dLLMs","ref_index":15,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JMSRQJSLUDDSPL2LTQ2TUBC53Q","json":"https://pith.science/pith/JMSRQJSLUDDSPL2LTQ2TUBC53Q.json","graph_json":"https://pith.science/api/pith-number/JMSRQJSLUDDSPL2LTQ2TUBC53Q/graph.json","events_json":"https://pith.science/api/pith-number/JMSRQJSLUDDSPL2LTQ2TUBC53Q/events.json","paper":"https://pith.science/paper/JMSRQJSL"},"agent_actions":{"view_html":"https://pith.science/pith/JMSRQJSLUDDSPL2LTQ2TUBC53Q","download_json":"https://pith.science/pith/JMSRQJSLUDDSPL2LTQ2TUBC53Q.json","view_paper":"https://pith.science/paper/JMSRQJSL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.14148&json=true","fetch_graph":"https://pith.science/api/pith-number/JMSRQJSLUDDSPL2LTQ2TUBC53Q/graph.json","fetch_events":"https://pith.science/api/pith-number/JMSRQJSLUDDSPL2LTQ2TUBC53Q/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JMSRQJSLUDDSPL2LTQ2TUBC53Q/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JMSRQJSLUDDSPL2LTQ2TUBC53Q/action/storage_attestation","attest_author":"https://pith.science/pith/JMSRQJSLUDDSPL2LTQ2TUBC53Q/action/author_attestation","sign_citation":"https://pith.science/pith/JMSRQJSLUDDSPL2LTQ2TUBC53Q/action/citation_signature","submit_replication":"https://pith.science/pith/JMSRQJSLUDDSPL2LTQ2TUBC53Q/action/replication_record"}},"created_at":"2026-07-05T11:57:59.201247+00:00","updated_at":"2026-07-05T11:57:59.201247+00:00"}