{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:A3HSD7RNO7OUBYY6QP6SW65AD3","short_pith_number":"pith:A3HSD7RN","canonical_record":{"source":{"id":"2605.22821","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-21T17:59:56Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"a6c484aac5b78ec4773ed021373eee170c199e2966d0a7338ebc647fb0ac9535","abstract_canon_sha256":"366d38082e892af7655de8fb2cef95d54279bdebaa44ab54ffff52af2c011a2f"},"schema_version":"1.0"},"canonical_sha256":"06cf21fe2d77dd40e31e83fd2b7ba01ed1152b4f0db83df720fbcd2c2914631f","source":{"kind":"arxiv","id":"2605.22821","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.22821","created_at":"2026-05-22T02:04:56Z"},{"alias_kind":"arxiv_version","alias_value":"2605.22821v1","created_at":"2026-05-22T02:04:56Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.22821","created_at":"2026-05-22T02:04:56Z"},{"alias_kind":"pith_short_12","alias_value":"A3HSD7RNO7OU","created_at":"2026-05-22T02:04:56Z"},{"alias_kind":"pith_short_16","alias_value":"A3HSD7RNO7OUBYY6","created_at":"2026-05-22T02:04:56Z"},{"alias_kind":"pith_short_8","alias_value":"A3HSD7RN","created_at":"2026-05-22T02:04:56Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:A3HSD7RNO7OUBYY6QP6SW65AD3","target":"record","payload":{"canonical_record":{"source":{"id":"2605.22821","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-21T17:59:56Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"a6c484aac5b78ec4773ed021373eee170c199e2966d0a7338ebc647fb0ac9535","abstract_canon_sha256":"366d38082e892af7655de8fb2cef95d54279bdebaa44ab54ffff52af2c011a2f"},"schema_version":"1.0"},"canonical_sha256":"06cf21fe2d77dd40e31e83fd2b7ba01ed1152b4f0db83df720fbcd2c2914631f","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T02:04:56.345741Z","signature_b64":"1QT1X0+nTB7rfscnochacOMM8dzkWo8QB4Xa6csj5GhouNzvQNB5M/xVs+/nYeO23deN+uGHABgFS4KAQ8/yAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"06cf21fe2d77dd40e31e83fd2b7ba01ed1152b4f0db83df720fbcd2c2914631f","last_reissued_at":"2026-05-22T02:04:56.345304Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T02:04:56.345304Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.22821","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-22T02:04:56Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"p5XPXRqLncxUjZ7jSFejZj2KxCm6jKqYGBJZD89CUbxU1ciOD6vqC8Hol5DwIjkFcLWMr65tZNMpGzx7AKnLCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-24T15:45:15.495093Z"},"content_sha256":"64666e92ce7c3a691a94d8d59c157caaade1478398288ed4d15ba633dfbaab2d","schema_version":"1.0","event_id":"sha256:64666e92ce7c3a691a94d8d59c157caaade1478398288ed4d15ba633dfbaab2d"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:A3HSD7RNO7OUBYY6QP6SW65AD3","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Tokenisation via Convex Relaxations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Craig W. Schmidt, Dennis Komm, Jan Tempus, Philip Whittington, Tiago Pimentel","submitted_at":"2026-05-21T17:59:56Z","abstract_excerpt":"Tokenisation is an integral part of the current NLP pipeline. Current tokenisation algorithms such as BPE and Unigram are greedy algorithms -- they make locally optimal decisions without considering the resulting vocabulary as a whole. We instead formulate tokeniser construction as a linear program and solve it using convex optimisation tools, yielding a new algorithm we call ConvexTok. We find ConvexTok consistently improves intrinsic tokenisation metrics and the bits-per-byte (BpB) achieved by language models; it also improves downstream task performance, but less consistently. Furthermore, "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.22821","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/2605.22821/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-22T02:04:56Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"qawKYXrUVGXK5BvTG6F1p8fJXgLidz2gh8OMO6Wjplef/wBWd4BFnfT012ViQBJsQsF4TrllDzWkvnSSlkc8Bw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-24T15:45:15.495862Z"},"content_sha256":"0944a1d69e03f802ef14b3b593a7d21fb1c8d0c3abdf99dee0667e292f1ae6cf","schema_version":"1.0","event_id":"sha256:0944a1d69e03f802ef14b3b593a7d21fb1c8d0c3abdf99dee0667e292f1ae6cf"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/A3HSD7RNO7OUBYY6QP6SW65AD3/bundle.json","state_url":"https://pith.science/pith/A3HSD7RNO7OUBYY6QP6SW65AD3/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/A3HSD7RNO7OUBYY6QP6SW65AD3/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-05-24T15:45:15Z","links":{"resolver":"https://pith.science/pith/A3HSD7RNO7OUBYY6QP6SW65AD3","bundle":"https://pith.science/pith/A3HSD7RNO7OUBYY6QP6SW65AD3/bundle.json","state":"https://pith.science/pith/A3HSD7RNO7OUBYY6QP6SW65AD3/state.json","well_known_bundle":"https://pith.science/.well-known/pith/A3HSD7RNO7OUBYY6QP6SW65AD3/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:A3HSD7RNO7OUBYY6QP6SW65AD3","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"366d38082e892af7655de8fb2cef95d54279bdebaa44ab54ffff52af2c011a2f","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-21T17:59:56Z","title_canon_sha256":"a6c484aac5b78ec4773ed021373eee170c199e2966d0a7338ebc647fb0ac9535"},"schema_version":"1.0","source":{"id":"2605.22821","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.22821","created_at":"2026-05-22T02:04:56Z"},{"alias_kind":"arxiv_version","alias_value":"2605.22821v1","created_at":"2026-05-22T02:04:56Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.22821","created_at":"2026-05-22T02:04:56Z"},{"alias_kind":"pith_short_12","alias_value":"A3HSD7RNO7OU","created_at":"2026-05-22T02:04:56Z"},{"alias_kind":"pith_short_16","alias_value":"A3HSD7RNO7OUBYY6","created_at":"2026-05-22T02:04:56Z"},{"alias_kind":"pith_short_8","alias_value":"A3HSD7RN","created_at":"2026-05-22T02:04:56Z"}],"graph_snapshots":[{"event_id":"sha256:0944a1d69e03f802ef14b3b593a7d21fb1c8d0c3abdf99dee0667e292f1ae6cf","target":"graph","created_at":"2026-05-22T02:04:56Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2605.22821/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Tokenisation is an integral part of the current NLP pipeline. Current tokenisation algorithms such as BPE and Unigram are greedy algorithms -- they make locally optimal decisions without considering the resulting vocabulary as a whole. We instead formulate tokeniser construction as a linear program and solve it using convex optimisation tools, yielding a new algorithm we call ConvexTok. We find ConvexTok consistently improves intrinsic tokenisation metrics and the bits-per-byte (BpB) achieved by language models; it also improves downstream task performance, but less consistently. Furthermore, ","authors_text":"Craig W. Schmidt, Dennis Komm, Jan Tempus, Philip Whittington, Tiago Pimentel","cross_cats":["cs.LG"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-21T17:59:56Z","title":"Tokenisation via Convex Relaxations"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.22821","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:64666e92ce7c3a691a94d8d59c157caaade1478398288ed4d15ba633dfbaab2d","target":"record","created_at":"2026-05-22T02:04:56Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"366d38082e892af7655de8fb2cef95d54279bdebaa44ab54ffff52af2c011a2f","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-21T17:59:56Z","title_canon_sha256":"a6c484aac5b78ec4773ed021373eee170c199e2966d0a7338ebc647fb0ac9535"},"schema_version":"1.0","source":{"id":"2605.22821","kind":"arxiv","version":1}},"canonical_sha256":"06cf21fe2d77dd40e31e83fd2b7ba01ed1152b4f0db83df720fbcd2c2914631f","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"06cf21fe2d77dd40e31e83fd2b7ba01ed1152b4f0db83df720fbcd2c2914631f","first_computed_at":"2026-05-22T02:04:56.345304Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-22T02:04:56.345304Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"1QT1X0+nTB7rfscnochacOMM8dzkWo8QB4Xa6csj5GhouNzvQNB5M/xVs+/nYeO23deN+uGHABgFS4KAQ8/yAg==","signature_status":"signed_v1","signed_at":"2026-05-22T02:04:56.345741Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.22821","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:64666e92ce7c3a691a94d8d59c157caaade1478398288ed4d15ba633dfbaab2d","sha256:0944a1d69e03f802ef14b3b593a7d21fb1c8d0c3abdf99dee0667e292f1ae6cf"],"state_sha256":"473f4ee0f93cbbda403442c0792791bfcbf14add2515ba443b81bace18b8ed14"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"grIRnmGw7GHfEBKIrPVT7WljtaP1SXPhdjR46/FbgHKGnQdm/hAcQU8H1GQtewqW3sd0yt/kGuXLG6d2LpynBg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-24T15:45:15.499540Z","bundle_sha256":"d830f734bd3eaebcf3c5f8ecacaabf1ec1bf36175de1090760a067ed2a0bfd73"}}