{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5J2ZX75S75N57ELCEK2ZOOOXAH","short_pith_number":"pith:5J2ZX75S","schema_version":"1.0","canonical_sha256":"ea759bffb2ff5bdf916222b59739d701e9ab3b0015bb98903da49ffaaa5cf054","source":{"kind":"arxiv","id":"2506.07541","version":1},"attestation_state":"computed","paper":{"title":"Bit-level BPE: Below the byte boundary","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Naoaki Okazaki, Sangwhan Moon, Tatsuya Hiraoka","submitted_at":"2025-06-09T08:28:16Z","abstract_excerpt":"Byte-level fallbacks for subword tokenization have become a common practice in large language models. In particular, it has been demonstrated to be incredibly effective as a pragmatic solution for preventing OOV, especially in the context of larger models. However, breaking a character down to individual bytes significantly increases the sequence length for long-tail tokens in languages such as Chinese, Japanese, and Korean (CJK) and other character-diverse contexts such as emoji. The increased sequence length results in longer computation during both training and inference. In this work, we p"},"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":"2506.07541","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-09T08:28:16Z","cross_cats_sorted":[],"title_canon_sha256":"c92bc11232c49aff5583a071705dd45110f7cc63166894ec54924f6fbff5bb56","abstract_canon_sha256":"16441cfab3a64f5bae67c5605640b7bd64d2b8fd9b8b9b00cfed11a97f4d8641"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:18:29.104709Z","signature_b64":"YU1mvyFtC7E8totfdyUI8MNpWepPFORytcU7X+H7O4gsFkbAVB/7RLa6L5guRtspw6A5BeCRI8dE5o+dLTzACQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ea759bffb2ff5bdf916222b59739d701e9ab3b0015bb98903da49ffaaa5cf054","last_reissued_at":"2026-07-05T11:18:29.104147Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:18:29.104147Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Bit-level BPE: Below the byte boundary","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Naoaki Okazaki, Sangwhan Moon, Tatsuya Hiraoka","submitted_at":"2025-06-09T08:28:16Z","abstract_excerpt":"Byte-level fallbacks for subword tokenization have become a common practice in large language models. In particular, it has been demonstrated to be incredibly effective as a pragmatic solution for preventing OOV, especially in the context of larger models. However, breaking a character down to individual bytes significantly increases the sequence length for long-tail tokens in languages such as Chinese, Japanese, and Korean (CJK) and other character-diverse contexts such as emoji. The increased sequence length results in longer computation during both training and inference. In this work, we p"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.07541","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/2506.07541/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":"2506.07541","created_at":"2026-07-05T11:18:29.104234+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.07541v1","created_at":"2026-07-05T11:18:29.104234+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.07541","created_at":"2026-07-05T11:18:29.104234+00:00"},{"alias_kind":"pith_short_12","alias_value":"5J2ZX75S75N5","created_at":"2026-07-05T11:18:29.104234+00:00"},{"alias_kind":"pith_short_16","alias_value":"5J2ZX75S75N57ELC","created_at":"2026-07-05T11:18:29.104234+00:00"},{"alias_kind":"pith_short_8","alias_value":"5J2ZX75S","created_at":"2026-07-05T11:18:29.104234+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2601.07220","citing_title":"The Roots of Performance Disparity in Multilingual Language Models: Intrinsic Modeling Difficulty or Design Choices?","ref_index":11,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5J2ZX75S75N57ELCEK2ZOOOXAH","json":"https://pith.science/pith/5J2ZX75S75N57ELCEK2ZOOOXAH.json","graph_json":"https://pith.science/api/pith-number/5J2ZX75S75N57ELCEK2ZOOOXAH/graph.json","events_json":"https://pith.science/api/pith-number/5J2ZX75S75N57ELCEK2ZOOOXAH/events.json","paper":"https://pith.science/paper/5J2ZX75S"},"agent_actions":{"view_html":"https://pith.science/pith/5J2ZX75S75N57ELCEK2ZOOOXAH","download_json":"https://pith.science/pith/5J2ZX75S75N57ELCEK2ZOOOXAH.json","view_paper":"https://pith.science/paper/5J2ZX75S","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.07541&json=true","fetch_graph":"https://pith.science/api/pith-number/5J2ZX75S75N57ELCEK2ZOOOXAH/graph.json","fetch_events":"https://pith.science/api/pith-number/5J2ZX75S75N57ELCEK2ZOOOXAH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5J2ZX75S75N57ELCEK2ZOOOXAH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5J2ZX75S75N57ELCEK2ZOOOXAH/action/storage_attestation","attest_author":"https://pith.science/pith/5J2ZX75S75N57ELCEK2ZOOOXAH/action/author_attestation","sign_citation":"https://pith.science/pith/5J2ZX75S75N57ELCEK2ZOOOXAH/action/citation_signature","submit_replication":"https://pith.science/pith/5J2ZX75S75N57ELCEK2ZOOOXAH/action/replication_record"}},"created_at":"2026-07-05T11:18:29.104234+00:00","updated_at":"2026-07-05T11:18:29.104234+00:00"}