{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:L7KQTA63VEFFF7ZXJQ4JIHAA2J","short_pith_number":"pith:L7KQTA63","schema_version":"1.0","canonical_sha256":"5fd50983dba90a52ff374c38941c00d276fb74591e94fa40496aba7f7288b5cf","source":{"kind":"arxiv","id":"2602.11133","version":2},"attestation_state":"computed","paper":{"title":"Just on Time: Token-Level Early Stopping for Diffusion Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Damian Rynczak, Dmytro Khamula, Jacek Ma{\\l}ecki, Michal Zakrzewski, Mykola Vysotskyi, Serhii Dmytryshyn, Severyn Shykula, Taras Rumezhak, Volodymyr Karpiv, Zakhar Kohut","submitted_at":"2026-02-11T18:44:04Z","abstract_excerpt":"Diffusion language models generate text through iterative refinement, a process that is often computationally inefficient because many tokens reach stability long before the final denoising step. We introduce a training-free, token-level early stopping approach that identifies convergence independently at each position. Our method leverages lightweight signals derived from the model's predictions and local context to dynamically determine when individual tokens can be finalized. This yields adaptive per-token freezing without task-specific fine-tuning, substantially reducing the total number o"},"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":"2602.11133","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-11T18:44:04Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"0e8128c7130b79ca1c32dd454be66cb4283bddee420797391bb64a270c16d553","abstract_canon_sha256":"ee3c919a04d17202a67195b5d7a5811a4561c96df8786a1b4dad65c994a7e55f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-04T01:59:10.700086Z","signature_b64":"tLcipe5xRfVgCI/DB/xkeYYcvnB63ecdbktDDfSND5c082BD6FR8Z/0K6Or7xshlcHKZ5rRikaFBL4FmzQPVBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5fd50983dba90a52ff374c38941c00d276fb74591e94fa40496aba7f7288b5cf","last_reissued_at":"2026-08-04T01:59:10.698469Z","signature_status":"signed_v1","first_computed_at":"2026-08-04T01:59:10.698469Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Just on Time: Token-Level Early Stopping for Diffusion Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Damian Rynczak, Dmytro Khamula, Jacek Ma{\\l}ecki, Michal Zakrzewski, Mykola Vysotskyi, Serhii Dmytryshyn, Severyn Shykula, Taras Rumezhak, Volodymyr Karpiv, Zakhar Kohut","submitted_at":"2026-02-11T18:44:04Z","abstract_excerpt":"Diffusion language models generate text through iterative refinement, a process that is often computationally inefficient because many tokens reach stability long before the final denoising step. We introduce a training-free, token-level early stopping approach that identifies convergence independently at each position. Our method leverages lightweight signals derived from the model's predictions and local context to dynamically determine when individual tokens can be finalized. This yields adaptive per-token freezing without task-specific fine-tuning, substantially reducing the total number o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.11133","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/2602.11133/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":"2602.11133","created_at":"2026-08-04T01:59:10.699841+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.11133v2","created_at":"2026-08-04T01:59:10.699841+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.11133","created_at":"2026-08-04T01:59:10.699841+00:00"},{"alias_kind":"pith_short_12","alias_value":"L7KQTA63VEFF","created_at":"2026-08-04T01:59:10.699841+00:00"},{"alias_kind":"pith_short_16","alias_value":"L7KQTA63VEFFF7ZX","created_at":"2026-08-04T01:59:10.699841+00:00"},{"alias_kind":"pith_short_8","alias_value":"L7KQTA63","created_at":"2026-08-04T01:59:10.699841+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2606.08810","citing_title":"Continuous Language Diffusion as a Decoder-Interface Problem","ref_index":30,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/L7KQTA63VEFFF7ZXJQ4JIHAA2J","json":"https://pith.science/pith/L7KQTA63VEFFF7ZXJQ4JIHAA2J.json","graph_json":"https://pith.science/api/pith-number/L7KQTA63VEFFF7ZXJQ4JIHAA2J/graph.json","events_json":"https://pith.science/api/pith-number/L7KQTA63VEFFF7ZXJQ4JIHAA2J/events.json","paper":"https://pith.science/paper/L7KQTA63"},"agent_actions":{"view_html":"https://pith.science/pith/L7KQTA63VEFFF7ZXJQ4JIHAA2J","download_json":"https://pith.science/pith/L7KQTA63VEFFF7ZXJQ4JIHAA2J.json","view_paper":"https://pith.science/paper/L7KQTA63","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.11133&json=true","fetch_graph":"https://pith.science/api/pith-number/L7KQTA63VEFFF7ZXJQ4JIHAA2J/graph.json","fetch_events":"https://pith.science/api/pith-number/L7KQTA63VEFFF7ZXJQ4JIHAA2J/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/L7KQTA63VEFFF7ZXJQ4JIHAA2J/action/timestamp_anchor","attest_storage":"https://pith.science/pith/L7KQTA63VEFFF7ZXJQ4JIHAA2J/action/storage_attestation","attest_author":"https://pith.science/pith/L7KQTA63VEFFF7ZXJQ4JIHAA2J/action/author_attestation","sign_citation":"https://pith.science/pith/L7KQTA63VEFFF7ZXJQ4JIHAA2J/action/citation_signature","submit_replication":"https://pith.science/pith/L7KQTA63VEFFF7ZXJQ4JIHAA2J/action/replication_record"}},"created_at":"2026-08-04T01:59:10.699841+00:00","updated_at":"2026-08-04T01:59:10.699841+00:00"}