{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:QZOS2VH2JDLI7CH54MKGGDCU5T","short_pith_number":"pith:QZOS2VH2","schema_version":"1.0","canonical_sha256":"865d2d54fa48d68f88fde314630c54ecdaf827dc67d79598e1b694fd006b1e01","source":{"kind":"arxiv","id":"2503.13423","version":3},"attestation_state":"computed","paper":{"title":"SuperBPE: Space Travel for Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Alisa Liu, Jonathan Hayase, Noah A. Smith, Sewoong Oh, Valentin Hofmann, Yejin Choi","submitted_at":"2025-03-17T17:53:23Z","abstract_excerpt":"The assumption across nearly all language model (LM) tokenization schemes is that tokens should be subwords, i.e., contained within word boundaries. While providing a seemingly reasonable inductive bias, is this common practice limiting the potential of modern LMs? Whitespace is not a reliable delimiter of meaning, as evidenced by multi-word expressions (e.g., \"by the way\"), crosslingual variation in the number of words needed to express a concept (e.g., \"spacesuit helmet\" in German is \"raumanzughelm\"), and languages that do not use whitespace at all (e.g., Chinese). To explore the potential 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":"2503.13423","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-03-17T17:53:23Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"e28cc7493ce4830af8649b777d34864c5f1de4b5794b40a6fd0bb4e44d47216d","abstract_canon_sha256":"269efe0b3110d365a989cbeef467753a0f0316020ce9b2a8167431d1bf6b73be"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:59:41.008179Z","signature_b64":"YbjAvNCugpADAcJXlEEI9MRrLaf0gr3i+ZfHVHRqKz7UkpdghSJGIhOa0B05VWc/NcZ35cH5VUvocSCCS4mBBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"865d2d54fa48d68f88fde314630c54ecdaf827dc67d79598e1b694fd006b1e01","last_reissued_at":"2026-07-05T11:59:41.007704Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:59:41.007704Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SuperBPE: Space Travel for Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Alisa Liu, Jonathan Hayase, Noah A. Smith, Sewoong Oh, Valentin Hofmann, Yejin Choi","submitted_at":"2025-03-17T17:53:23Z","abstract_excerpt":"The assumption across nearly all language model (LM) tokenization schemes is that tokens should be subwords, i.e., contained within word boundaries. While providing a seemingly reasonable inductive bias, is this common practice limiting the potential of modern LMs? Whitespace is not a reliable delimiter of meaning, as evidenced by multi-word expressions (e.g., \"by the way\"), crosslingual variation in the number of words needed to express a concept (e.g., \"spacesuit helmet\" in German is \"raumanzughelm\"), and languages that do not use whitespace at all (e.g., Chinese). To explore the potential o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.13423","kind":"arxiv","version":3},"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/2503.13423/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":"2503.13423","created_at":"2026-07-05T11:59:41.007760+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.13423v3","created_at":"2026-07-05T11:59:41.007760+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.13423","created_at":"2026-07-05T11:59:41.007760+00:00"},{"alias_kind":"pith_short_12","alias_value":"QZOS2VH2JDLI","created_at":"2026-07-05T11:59:41.007760+00:00"},{"alias_kind":"pith_short_16","alias_value":"QZOS2VH2JDLI7CH5","created_at":"2026-07-05T11:59:41.007760+00:00"},{"alias_kind":"pith_short_8","alias_value":"QZOS2VH2","created_at":"2026-07-05T11:59:41.007760+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.05691","citing_title":"Where to cut, how deep: BPE and Unigram-LM on chemistry SMILES","ref_index":29,"is_internal_anchor":true},{"citing_arxiv_id":"2605.29379","citing_title":"BrahmicTokenizer-131K: An Indic-Capable Drop-In Replacement for o200k_base","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2506.14123","citing_title":"Sampling from Your Language Model One Byte at a Time","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2510.17001","citing_title":"Vocab Diet: Reshaping the Vocabulary of LLMs via Vector Arithmetic","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2603.06351","citing_title":"DC-DiT: Adaptive Compute and Elastic Inference for Visual Generation via Dynamic Chunking","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01188","citing_title":"Compute Optimal Tokenization","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2604.05192","citing_title":"Faster Superword Tokenization","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QZOS2VH2JDLI7CH54MKGGDCU5T","json":"https://pith.science/pith/QZOS2VH2JDLI7CH54MKGGDCU5T.json","graph_json":"https://pith.science/api/pith-number/QZOS2VH2JDLI7CH54MKGGDCU5T/graph.json","events_json":"https://pith.science/api/pith-number/QZOS2VH2JDLI7CH54MKGGDCU5T/events.json","paper":"https://pith.science/paper/QZOS2VH2"},"agent_actions":{"view_html":"https://pith.science/pith/QZOS2VH2JDLI7CH54MKGGDCU5T","download_json":"https://pith.science/pith/QZOS2VH2JDLI7CH54MKGGDCU5T.json","view_paper":"https://pith.science/paper/QZOS2VH2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.13423&json=true","fetch_graph":"https://pith.science/api/pith-number/QZOS2VH2JDLI7CH54MKGGDCU5T/graph.json","fetch_events":"https://pith.science/api/pith-number/QZOS2VH2JDLI7CH54MKGGDCU5T/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QZOS2VH2JDLI7CH54MKGGDCU5T/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QZOS2VH2JDLI7CH54MKGGDCU5T/action/storage_attestation","attest_author":"https://pith.science/pith/QZOS2VH2JDLI7CH54MKGGDCU5T/action/author_attestation","sign_citation":"https://pith.science/pith/QZOS2VH2JDLI7CH54MKGGDCU5T/action/citation_signature","submit_replication":"https://pith.science/pith/QZOS2VH2JDLI7CH54MKGGDCU5T/action/replication_record"}},"created_at":"2026-07-05T11:59:41.007760+00:00","updated_at":"2026-07-05T11:59:41.007760+00:00"}