{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:GUZ6GPMARS54F5A5MQQXOB6UG3","short_pith_number":"pith:GUZ6GPMA","schema_version":"1.0","canonical_sha256":"3533e33d808cbbc2f41d64217707d436ce8d6d9947c410933ab2f57d19e0ed3d","source":{"kind":"arxiv","id":"2402.18376","version":2},"attestation_state":"computed","paper":{"title":"Tokenization Is More Than Compression","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Alec Alameddine, Chris Tanner, Craig W. Schmidt, Haoran Zhang, Omri Uzan, Varshini Reddy, Yuval Pinter","submitted_at":"2024-02-28T14:52:15Z","abstract_excerpt":"Tokenization is a foundational step in natural language processing (NLP) tasks, bridging raw text and language models. Existing tokenization approaches like Byte-Pair Encoding (BPE) originate from the field of data compression, and it has been suggested that the effectiveness of BPE stems from its ability to condense text into a relatively small number of tokens. We test the hypothesis that fewer tokens lead to better downstream performance by introducing PathPiece, a new tokenizer that segments a document's text into the minimum number of tokens for a given vocabulary. Through extensive exper"},"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":"2402.18376","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2024-02-28T14:52:15Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"a990694e92e9cfd156ba74811cedf498751c13aabb5f0143c07a96a3f0dfd136","abstract_canon_sha256":"18581e685ff17d5e010a8b11a265a72b962d2167490c13a43b1af04e5ca13474"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:16:46.688843Z","signature_b64":"7Oo2RxL5vjNg99A+B+HdtSf+gyCQN6W8R1CdSDZ4CGC1swcBmCBCP7Aaavbq8Un/bEAMivro0SwM21qlQOMkDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3533e33d808cbbc2f41d64217707d436ce8d6d9947c410933ab2f57d19e0ed3d","last_reissued_at":"2026-07-05T09:16:46.688343Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:16:46.688343Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Tokenization Is More Than Compression","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Alec Alameddine, Chris Tanner, Craig W. Schmidt, Haoran Zhang, Omri Uzan, Varshini Reddy, Yuval Pinter","submitted_at":"2024-02-28T14:52:15Z","abstract_excerpt":"Tokenization is a foundational step in natural language processing (NLP) tasks, bridging raw text and language models. Existing tokenization approaches like Byte-Pair Encoding (BPE) originate from the field of data compression, and it has been suggested that the effectiveness of BPE stems from its ability to condense text into a relatively small number of tokens. We test the hypothesis that fewer tokens lead to better downstream performance by introducing PathPiece, a new tokenizer that segments a document's text into the minimum number of tokens for a given vocabulary. Through extensive exper"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.18376","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/2402.18376/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":"2402.18376","created_at":"2026-07-05T09:16:46.688404+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.18376v2","created_at":"2026-07-05T09:16:46.688404+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.18376","created_at":"2026-07-05T09:16:46.688404+00:00"},{"alias_kind":"pith_short_12","alias_value":"GUZ6GPMARS54","created_at":"2026-07-05T09:16:46.688404+00:00"},{"alias_kind":"pith_short_16","alias_value":"GUZ6GPMARS54F5A5","created_at":"2026-07-05T09:16:46.688404+00:00"},{"alias_kind":"pith_short_8","alias_value":"GUZ6GPMA","created_at":"2026-07-05T09:16:46.688404+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.08713","citing_title":"The price of incrementality in k-center clustering","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2601.07220","citing_title":"The Roots of Performance Disparity in Multilingual Language Models: Intrinsic Modeling Difficulty or Design Choices?","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2602.04289","citing_title":"Proxy Compression for Language Modeling","ref_index":11,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GUZ6GPMARS54F5A5MQQXOB6UG3","json":"https://pith.science/pith/GUZ6GPMARS54F5A5MQQXOB6UG3.json","graph_json":"https://pith.science/api/pith-number/GUZ6GPMARS54F5A5MQQXOB6UG3/graph.json","events_json":"https://pith.science/api/pith-number/GUZ6GPMARS54F5A5MQQXOB6UG3/events.json","paper":"https://pith.science/paper/GUZ6GPMA"},"agent_actions":{"view_html":"https://pith.science/pith/GUZ6GPMARS54F5A5MQQXOB6UG3","download_json":"https://pith.science/pith/GUZ6GPMARS54F5A5MQQXOB6UG3.json","view_paper":"https://pith.science/paper/GUZ6GPMA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.18376&json=true","fetch_graph":"https://pith.science/api/pith-number/GUZ6GPMARS54F5A5MQQXOB6UG3/graph.json","fetch_events":"https://pith.science/api/pith-number/GUZ6GPMARS54F5A5MQQXOB6UG3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GUZ6GPMARS54F5A5MQQXOB6UG3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GUZ6GPMARS54F5A5MQQXOB6UG3/action/storage_attestation","attest_author":"https://pith.science/pith/GUZ6GPMARS54F5A5MQQXOB6UG3/action/author_attestation","sign_citation":"https://pith.science/pith/GUZ6GPMARS54F5A5MQQXOB6UG3/action/citation_signature","submit_replication":"https://pith.science/pith/GUZ6GPMARS54F5A5MQQXOB6UG3/action/replication_record"}},"created_at":"2026-07-05T09:16:46.688404+00:00","updated_at":"2026-07-05T09:16:46.688404+00:00"}