{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:KGABQLFYPGB34NPDHJEEY2U5J3","short_pith_number":"pith:KGABQLFY","schema_version":"1.0","canonical_sha256":"5180182cb87983be35e33a484c6a9d4efbbcfe7414d573147a672c7c36bbc629","source":{"kind":"arxiv","id":"2507.16323","version":1},"attestation_state":"computed","paper":{"title":"SpeLLM: Character-Level Multi-Head Decoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Amit Ben-Artzy, Roy Schwartz","submitted_at":"2025-07-22T08:07:06Z","abstract_excerpt":"Scaling LLM vocabulary is often used to reduce input sequence length and alleviate attention's quadratic cost. Yet, current LLM architectures impose a critical bottleneck to this procedure: the output projection layer scales linearly with vocabulary size, rendering substantial expansion impractical. We propose SpeLLM, a method that decouples input and output vocabularies by predicting character-level strings through multiple output heads. In SpeLLM, each of the $k$ linear heads predicts a single character simultaneously, enabling the model to represent a much larger output space using smaller,"},"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":"2507.16323","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-07-22T08:07:06Z","cross_cats_sorted":[],"title_canon_sha256":"96cfcb285033e57d9a256ebe206239af4f8ee76ad49c509d6a8af31f02604029","abstract_canon_sha256":"7f46883149510fa04d858296e4bb660c4ddc6a7192a95a64dcbd99141b4a9a07"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:41:07.563503Z","signature_b64":"ZlNpIRL/RNs5sGI+kmdoEfmSr2aVf9tfgDxhVznIBDuEFCYq/wfzeNEvRgL0/SN9WMBHNHL9WAJ1Yszg3eakDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5180182cb87983be35e33a484c6a9d4efbbcfe7414d573147a672c7c36bbc629","last_reissued_at":"2026-07-05T11:41:07.563074Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:41:07.563074Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SpeLLM: Character-Level Multi-Head Decoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Amit Ben-Artzy, Roy Schwartz","submitted_at":"2025-07-22T08:07:06Z","abstract_excerpt":"Scaling LLM vocabulary is often used to reduce input sequence length and alleviate attention's quadratic cost. Yet, current LLM architectures impose a critical bottleneck to this procedure: the output projection layer scales linearly with vocabulary size, rendering substantial expansion impractical. We propose SpeLLM, a method that decouples input and output vocabularies by predicting character-level strings through multiple output heads. In SpeLLM, each of the $k$ linear heads predicts a single character simultaneously, enabling the model to represent a much larger output space using smaller,"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.16323","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/2507.16323/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":"2507.16323","created_at":"2026-07-05T11:41:07.563124+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.16323v1","created_at":"2026-07-05T11:41:07.563124+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.16323","created_at":"2026-07-05T11:41:07.563124+00:00"},{"alias_kind":"pith_short_12","alias_value":"KGABQLFYPGB3","created_at":"2026-07-05T11:41:07.563124+00:00"},{"alias_kind":"pith_short_16","alias_value":"KGABQLFYPGB34NPD","created_at":"2026-07-05T11:41:07.563124+00:00"},{"alias_kind":"pith_short_8","alias_value":"KGABQLFY","created_at":"2026-07-05T11:41:07.563124+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/KGABQLFYPGB34NPDHJEEY2U5J3","json":"https://pith.science/pith/KGABQLFYPGB34NPDHJEEY2U5J3.json","graph_json":"https://pith.science/api/pith-number/KGABQLFYPGB34NPDHJEEY2U5J3/graph.json","events_json":"https://pith.science/api/pith-number/KGABQLFYPGB34NPDHJEEY2U5J3/events.json","paper":"https://pith.science/paper/KGABQLFY"},"agent_actions":{"view_html":"https://pith.science/pith/KGABQLFYPGB34NPDHJEEY2U5J3","download_json":"https://pith.science/pith/KGABQLFYPGB34NPDHJEEY2U5J3.json","view_paper":"https://pith.science/paper/KGABQLFY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.16323&json=true","fetch_graph":"https://pith.science/api/pith-number/KGABQLFYPGB34NPDHJEEY2U5J3/graph.json","fetch_events":"https://pith.science/api/pith-number/KGABQLFYPGB34NPDHJEEY2U5J3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KGABQLFYPGB34NPDHJEEY2U5J3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KGABQLFYPGB34NPDHJEEY2U5J3/action/storage_attestation","attest_author":"https://pith.science/pith/KGABQLFYPGB34NPDHJEEY2U5J3/action/author_attestation","sign_citation":"https://pith.science/pith/KGABQLFYPGB34NPDHJEEY2U5J3/action/citation_signature","submit_replication":"https://pith.science/pith/KGABQLFYPGB34NPDHJEEY2U5J3/action/replication_record"}},"created_at":"2026-07-05T11:41:07.563124+00:00","updated_at":"2026-07-05T11:41:07.563124+00:00"}