{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:56QGBLRWXTTZ3LBZQQW2KPNZMT","short_pith_number":"pith:56QGBLRW","schema_version":"1.0","canonical_sha256":"efa060ae36bce79dac39842da53db964f839cb47173b96648c0e6751a60cedeb","source":{"kind":"arxiv","id":"2306.16837","version":3},"attestation_state":"computed","paper":{"title":"A Formal Perspective on Byte-Pair Encoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["math.OC"],"primary_cat":"cs.CL","authors_text":"Clara Meister, Juan Luis Gastaldi, Li Du, Mrinmaya Sachan, Ryan Cotterell, Tim Vieira, Vil\\'em Zouhar","submitted_at":"2023-06-29T10:29:23Z","abstract_excerpt":"Byte-Pair Encoding (BPE) is a popular algorithm used for tokenizing data in NLP, despite being devised initially as a compression method. BPE appears to be a greedy algorithm at face value, but the underlying optimization problem that BPE seeks to solve has not yet been laid down. We formalize BPE as a combinatorial optimization problem. Via submodular functions, we prove that the iterative greedy version is a $\\frac{1}{{\\sigma(\\boldsymbol{\\mu}^\\star)}}(1-e^{-{\\sigma(\\boldsymbol{\\mu}^\\star)}})$-approximation of an optimal merge sequence, where ${\\sigma(\\boldsymbol{\\mu}^\\star)}$ is the total ba"},"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":"2306.16837","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-06-29T10:29:23Z","cross_cats_sorted":["math.OC"],"title_canon_sha256":"69219e5e111c02cbd81019f415ab4c5239cc1e3b354de244db68e08e8fd9af70","abstract_canon_sha256":"9bd467ccf262c5951852ecc1cf90d15739640b26435e207cb8f4a9260e302941"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:01:24.901445Z","signature_b64":"P/7MMc6tcb2EaKwBoV6Fu1J9FmUDXYrdwO8mPk7IF1Bz62JPi8Mb3P5iWxnP0WNfdm2kY/UHwSTCd5NoECiiDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"efa060ae36bce79dac39842da53db964f839cb47173b96648c0e6751a60cedeb","last_reissued_at":"2026-07-05T09:01:24.900960Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:01:24.900960Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Formal Perspective on Byte-Pair Encoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["math.OC"],"primary_cat":"cs.CL","authors_text":"Clara Meister, Juan Luis Gastaldi, Li Du, Mrinmaya Sachan, Ryan Cotterell, Tim Vieira, Vil\\'em Zouhar","submitted_at":"2023-06-29T10:29:23Z","abstract_excerpt":"Byte-Pair Encoding (BPE) is a popular algorithm used for tokenizing data in NLP, despite being devised initially as a compression method. BPE appears to be a greedy algorithm at face value, but the underlying optimization problem that BPE seeks to solve has not yet been laid down. We formalize BPE as a combinatorial optimization problem. Via submodular functions, we prove that the iterative greedy version is a $\\frac{1}{{\\sigma(\\boldsymbol{\\mu}^\\star)}}(1-e^{-{\\sigma(\\boldsymbol{\\mu}^\\star)}})$-approximation of an optimal merge sequence, where ${\\sigma(\\boldsymbol{\\mu}^\\star)}$ is the total ba"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.16837","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/2306.16837/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":"2306.16837","created_at":"2026-07-05T09:01:24.901029+00:00"},{"alias_kind":"arxiv_version","alias_value":"2306.16837v3","created_at":"2026-07-05T09:01:24.901029+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.16837","created_at":"2026-07-05T09:01:24.901029+00:00"},{"alias_kind":"pith_short_12","alias_value":"56QGBLRWXTTZ","created_at":"2026-07-05T09:01:24.901029+00:00"},{"alias_kind":"pith_short_16","alias_value":"56QGBLRWXTTZ3LBZ","created_at":"2026-07-05T09:01:24.901029+00:00"},{"alias_kind":"pith_short_8","alias_value":"56QGBLRW","created_at":"2026-07-05T09:01:24.901029+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2411.12240","citing_title":"Evaluating Tokenizer Performance of Large Language Models Across Official Indian Languages","ref_index":8,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/56QGBLRWXTTZ3LBZQQW2KPNZMT","json":"https://pith.science/pith/56QGBLRWXTTZ3LBZQQW2KPNZMT.json","graph_json":"https://pith.science/api/pith-number/56QGBLRWXTTZ3LBZQQW2KPNZMT/graph.json","events_json":"https://pith.science/api/pith-number/56QGBLRWXTTZ3LBZQQW2KPNZMT/events.json","paper":"https://pith.science/paper/56QGBLRW"},"agent_actions":{"view_html":"https://pith.science/pith/56QGBLRWXTTZ3LBZQQW2KPNZMT","download_json":"https://pith.science/pith/56QGBLRWXTTZ3LBZQQW2KPNZMT.json","view_paper":"https://pith.science/paper/56QGBLRW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2306.16837&json=true","fetch_graph":"https://pith.science/api/pith-number/56QGBLRWXTTZ3LBZQQW2KPNZMT/graph.json","fetch_events":"https://pith.science/api/pith-number/56QGBLRWXTTZ3LBZQQW2KPNZMT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/56QGBLRWXTTZ3LBZQQW2KPNZMT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/56QGBLRWXTTZ3LBZQQW2KPNZMT/action/storage_attestation","attest_author":"https://pith.science/pith/56QGBLRWXTTZ3LBZQQW2KPNZMT/action/author_attestation","sign_citation":"https://pith.science/pith/56QGBLRWXTTZ3LBZQQW2KPNZMT/action/citation_signature","submit_replication":"https://pith.science/pith/56QGBLRWXTTZ3LBZQQW2KPNZMT/action/replication_record"}},"created_at":"2026-07-05T09:01:24.901029+00:00","updated_at":"2026-07-05T09:01:24.901029+00:00"}