{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:F33FWQSSYV2RX4XXHSAW5RZTF6","short_pith_number":"pith:F33FWQSS","schema_version":"1.0","canonical_sha256":"2ef65b4252c5751bf2f73c816ec7332fbd7d3a295d832d5774a6aace7c27f0e4","source":{"kind":"arxiv","id":"2605.22471","version":1},"attestation_state":"computed","paper":{"title":"Lost in Tokenization: Fundamental Trade-offs in Graph Tokenization for Transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Amir Globerson, Clayton Sanford, Gilad Yehudai, Gil Harari, Joan Bruna, Maya Bechler-Speicher","submitted_at":"2026-05-21T13:32:20Z","abstract_excerpt":"Transformers have become a central architecture for graph learning, but their application to graphs requires first choosing a tokenization: a graph-to-token map that determines which structural information is exposed at the input. In this work, we show that this choice is a fundamental component of transformer expressivity. We examine three tokenizations that serve as building blocks for many existing graph tokenizations: spectral, random-walk, and adjacency tokenizations. We prove that different tokenizations induce distinct depth regimes: the same graph computation may be realizable by a sha"},"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":"2605.22471","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-21T13:32:20Z","cross_cats_sorted":[],"title_canon_sha256":"8c7e358a905c4438a83632d5fafd4073a63620a21a7958d5de5fe9d2dd4d4f05","abstract_canon_sha256":"d88a4343b5f3e78e780dca475da6260cff0947579a0b70cb140b127e232f7e5f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T01:04:44.507052Z","signature_b64":"gLitU806m0l24qOP6IGDUR6zsPcZlAODjjag0i9fXW6Qr/QgBJdN/6xdcTRnsob10AG0+gVuyQX5elztnCQCDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2ef65b4252c5751bf2f73c816ec7332fbd7d3a295d832d5774a6aace7c27f0e4","last_reissued_at":"2026-05-22T01:04:44.506492Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T01:04:44.506492Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Lost in Tokenization: Fundamental Trade-offs in Graph Tokenization for Transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Amir Globerson, Clayton Sanford, Gilad Yehudai, Gil Harari, Joan Bruna, Maya Bechler-Speicher","submitted_at":"2026-05-21T13:32:20Z","abstract_excerpt":"Transformers have become a central architecture for graph learning, but their application to graphs requires first choosing a tokenization: a graph-to-token map that determines which structural information is exposed at the input. In this work, we show that this choice is a fundamental component of transformer expressivity. We examine three tokenizations that serve as building blocks for many existing graph tokenizations: spectral, random-walk, and adjacency tokenizations. We prove that different tokenizations induce distinct depth regimes: the same graph computation may be realizable by a sha"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.22471","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/2605.22471/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":"2605.22471","created_at":"2026-05-22T01:04:44.506592+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.22471v1","created_at":"2026-05-22T01:04:44.506592+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.22471","created_at":"2026-05-22T01:04:44.506592+00:00"},{"alias_kind":"pith_short_12","alias_value":"F33FWQSSYV2R","created_at":"2026-05-22T01:04:44.506592+00:00"},{"alias_kind":"pith_short_16","alias_value":"F33FWQSSYV2RX4XX","created_at":"2026-05-22T01:04:44.506592+00:00"},{"alias_kind":"pith_short_8","alias_value":"F33FWQSS","created_at":"2026-05-22T01:04:44.506592+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/F33FWQSSYV2RX4XXHSAW5RZTF6","json":"https://pith.science/pith/F33FWQSSYV2RX4XXHSAW5RZTF6.json","graph_json":"https://pith.science/api/pith-number/F33FWQSSYV2RX4XXHSAW5RZTF6/graph.json","events_json":"https://pith.science/api/pith-number/F33FWQSSYV2RX4XXHSAW5RZTF6/events.json","paper":"https://pith.science/paper/F33FWQSS"},"agent_actions":{"view_html":"https://pith.science/pith/F33FWQSSYV2RX4XXHSAW5RZTF6","download_json":"https://pith.science/pith/F33FWQSSYV2RX4XXHSAW5RZTF6.json","view_paper":"https://pith.science/paper/F33FWQSS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.22471&json=true","fetch_graph":"https://pith.science/api/pith-number/F33FWQSSYV2RX4XXHSAW5RZTF6/graph.json","fetch_events":"https://pith.science/api/pith-number/F33FWQSSYV2RX4XXHSAW5RZTF6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/F33FWQSSYV2RX4XXHSAW5RZTF6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/F33FWQSSYV2RX4XXHSAW5RZTF6/action/storage_attestation","attest_author":"https://pith.science/pith/F33FWQSSYV2RX4XXHSAW5RZTF6/action/author_attestation","sign_citation":"https://pith.science/pith/F33FWQSSYV2RX4XXHSAW5RZTF6/action/citation_signature","submit_replication":"https://pith.science/pith/F33FWQSSYV2RX4XXHSAW5RZTF6/action/replication_record"}},"created_at":"2026-05-22T01:04:44.506592+00:00","updated_at":"2026-05-22T01:04:44.506592+00:00"}