{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:LJEO4HHJLYGYEVZROBPZV7TLMX","short_pith_number":"pith:LJEO4HHJ","schema_version":"1.0","canonical_sha256":"5a48ee1ce95e0d825731705f9afe6b65daae3e6adec477616c2a87de4c8f208d","source":{"kind":"arxiv","id":"2506.07956","version":1},"attestation_state":"computed","paper":{"title":"Language Models over Canonical Byte-Pair Encodings","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.FL","cs.LG"],"primary_cat":"cs.CL","authors_text":"Benjamin Lebrun, Brian DuSell, Clemente Pasti, Juan Luis Gastaldi, Mario Giulianelli, Ryan Cotterell, Tianyu Liu, Timothy J. O'Donnell, Tim Vieira, Yahya Emara","submitted_at":"2025-06-09T17:26:14Z","abstract_excerpt":"Modern language models represent probability distributions over character strings as distributions over (shorter) token strings derived via a deterministic tokenizer, such as byte-pair encoding. While this approach is highly effective at scaling up language models to large corpora, its current incarnations have a concerning property: the model assigns nonzero probability mass to an exponential number of $\\it{noncanonical}$ token encodings of each character string -- these are token strings that decode to valid character strings but are impossible under the deterministic tokenizer (i.e., they w"},"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":"2506.07956","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-09T17:26:14Z","cross_cats_sorted":["cs.FL","cs.LG"],"title_canon_sha256":"34b89823c76c1d2556551c84f06cb8e86c05d142e6878586a2c2acf89661bea5","abstract_canon_sha256":"c5909f5b7ce1981d10250c807e33fbffedcdc7756ca33dc26c85a28907341181"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:18:36.033686Z","signature_b64":"MyyrDoYyyjW42LI3zq+A6RKKQbEb8k+gXjgQksmB/WaLoRBKevMmqwkpYHMCDNsYlM7sCLmnG6w14FwaZ+VQBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5a48ee1ce95e0d825731705f9afe6b65daae3e6adec477616c2a87de4c8f208d","last_reissued_at":"2026-07-05T11:18:36.033126Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:18:36.033126Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Language Models over Canonical Byte-Pair Encodings","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.FL","cs.LG"],"primary_cat":"cs.CL","authors_text":"Benjamin Lebrun, Brian DuSell, Clemente Pasti, Juan Luis Gastaldi, Mario Giulianelli, Ryan Cotterell, Tianyu Liu, Timothy J. O'Donnell, Tim Vieira, Yahya Emara","submitted_at":"2025-06-09T17:26:14Z","abstract_excerpt":"Modern language models represent probability distributions over character strings as distributions over (shorter) token strings derived via a deterministic tokenizer, such as byte-pair encoding. While this approach is highly effective at scaling up language models to large corpora, its current incarnations have a concerning property: the model assigns nonzero probability mass to an exponential number of $\\it{noncanonical}$ token encodings of each character string -- these are token strings that decode to valid character strings but are impossible under the deterministic tokenizer (i.e., they w"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.07956","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/2506.07956/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":"2506.07956","created_at":"2026-07-05T11:18:36.033190+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.07956v1","created_at":"2026-07-05T11:18:36.033190+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.07956","created_at":"2026-07-05T11:18:36.033190+00:00"},{"alias_kind":"pith_short_12","alias_value":"LJEO4HHJLYGY","created_at":"2026-07-05T11:18:36.033190+00:00"},{"alias_kind":"pith_short_16","alias_value":"LJEO4HHJLYGYEVZR","created_at":"2026-07-05T11:18:36.033190+00:00"},{"alias_kind":"pith_short_8","alias_value":"LJEO4HHJ","created_at":"2026-07-05T11:18:36.033190+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2506.14123","citing_title":"Sampling from Your Language Model One Byte at a Time","ref_index":76,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LJEO4HHJLYGYEVZROBPZV7TLMX","json":"https://pith.science/pith/LJEO4HHJLYGYEVZROBPZV7TLMX.json","graph_json":"https://pith.science/api/pith-number/LJEO4HHJLYGYEVZROBPZV7TLMX/graph.json","events_json":"https://pith.science/api/pith-number/LJEO4HHJLYGYEVZROBPZV7TLMX/events.json","paper":"https://pith.science/paper/LJEO4HHJ"},"agent_actions":{"view_html":"https://pith.science/pith/LJEO4HHJLYGYEVZROBPZV7TLMX","download_json":"https://pith.science/pith/LJEO4HHJLYGYEVZROBPZV7TLMX.json","view_paper":"https://pith.science/paper/LJEO4HHJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.07956&json=true","fetch_graph":"https://pith.science/api/pith-number/LJEO4HHJLYGYEVZROBPZV7TLMX/graph.json","fetch_events":"https://pith.science/api/pith-number/LJEO4HHJLYGYEVZROBPZV7TLMX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LJEO4HHJLYGYEVZROBPZV7TLMX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LJEO4HHJLYGYEVZROBPZV7TLMX/action/storage_attestation","attest_author":"https://pith.science/pith/LJEO4HHJLYGYEVZROBPZV7TLMX/action/author_attestation","sign_citation":"https://pith.science/pith/LJEO4HHJLYGYEVZROBPZV7TLMX/action/citation_signature","submit_replication":"https://pith.science/pith/LJEO4HHJLYGYEVZROBPZV7TLMX/action/replication_record"}},"created_at":"2026-07-05T11:18:36.033190+00:00","updated_at":"2026-07-05T11:18:36.033190+00:00"}