{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:K7DCIYGEQ2LWEJ22Q65EKL4U5J","short_pith_number":"pith:K7DCIYGE","schema_version":"1.0","canonical_sha256":"57c62460c4869762275a87ba452f94ea622237297c6dc1d3c7cc919f3b5deaf0","source":{"kind":"arxiv","id":"2605.26444","version":1},"attestation_state":"computed","paper":{"title":"MicroSpec: Accelerating Speculative Decoding with Lightweight In-Context Vocabularies","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chenghua Wang, Daliang Xu, Mengwei Xu, Yinyuan Zhang, Yun Ma, Zhiyang Chen","submitted_at":"2026-04-08T05:33:40Z","abstract_excerpt":"Large language models typically employ vocabularies of over 100k tokens, which creates a major computational bottleneck at the final linear projection layer when performing speculative decoding. Current methods for vocabulary pruning depend on either fixed or coarse-grained sub-vocabularies, requiring around 30k active tokens to preserve the quality of the draft model. We introduce MicroSpec, a training-free technique that overcomes this limitation by building a compact, context-sensitive active vocabulary on the fly for every decoding step. Exploiting the natural temporal locality found in la"},"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":"2605.26444","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-04-08T05:33:40Z","cross_cats_sorted":[],"title_canon_sha256":"b22349946715014339c697c4fe5f10c7845d3219ae96e328c85d7559a622d202","abstract_canon_sha256":"6d7385d4f95796fc5846bcf37ff26e8aa041b271a475d019ad1b5dc6640ccf3c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-27T01:05:18.252748Z","signature_b64":"f6vZ1kXQBvTXpvna9wDqJgOUOdZ9yE6Eoc0p6WcWU6yk4bEkrAxwmExBMn5r3Bk38sXINWkYg03flPTkUPNuAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"57c62460c4869762275a87ba452f94ea622237297c6dc1d3c7cc919f3b5deaf0","last_reissued_at":"2026-05-27T01:05:18.251961Z","signature_status":"signed_v1","first_computed_at":"2026-05-27T01:05:18.251961Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MicroSpec: Accelerating Speculative Decoding with Lightweight In-Context Vocabularies","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chenghua Wang, Daliang Xu, Mengwei Xu, Yinyuan Zhang, Yun Ma, Zhiyang Chen","submitted_at":"2026-04-08T05:33:40Z","abstract_excerpt":"Large language models typically employ vocabularies of over 100k tokens, which creates a major computational bottleneck at the final linear projection layer when performing speculative decoding. Current methods for vocabulary pruning depend on either fixed or coarse-grained sub-vocabularies, requiring around 30k active tokens to preserve the quality of the draft model. We introduce MicroSpec, a training-free technique that overcomes this limitation by building a compact, context-sensitive active vocabulary on the fly for every decoding step. Exploiting the natural temporal locality found in la"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.26444","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/2605.26444/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":"2605.26444","created_at":"2026-05-27T01:05:18.252093+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.26444v1","created_at":"2026-05-27T01:05:18.252093+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.26444","created_at":"2026-05-27T01:05:18.252093+00:00"},{"alias_kind":"pith_short_12","alias_value":"K7DCIYGEQ2LW","created_at":"2026-05-27T01:05:18.252093+00:00"},{"alias_kind":"pith_short_16","alias_value":"K7DCIYGEQ2LWEJ22","created_at":"2026-05-27T01:05:18.252093+00:00"},{"alias_kind":"pith_short_8","alias_value":"K7DCIYGE","created_at":"2026-05-27T01:05:18.252093+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/K7DCIYGEQ2LWEJ22Q65EKL4U5J","json":"https://pith.science/pith/K7DCIYGEQ2LWEJ22Q65EKL4U5J.json","graph_json":"https://pith.science/api/pith-number/K7DCIYGEQ2LWEJ22Q65EKL4U5J/graph.json","events_json":"https://pith.science/api/pith-number/K7DCIYGEQ2LWEJ22Q65EKL4U5J/events.json","paper":"https://pith.science/paper/K7DCIYGE"},"agent_actions":{"view_html":"https://pith.science/pith/K7DCIYGEQ2LWEJ22Q65EKL4U5J","download_json":"https://pith.science/pith/K7DCIYGEQ2LWEJ22Q65EKL4U5J.json","view_paper":"https://pith.science/paper/K7DCIYGE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.26444&json=true","fetch_graph":"https://pith.science/api/pith-number/K7DCIYGEQ2LWEJ22Q65EKL4U5J/graph.json","fetch_events":"https://pith.science/api/pith-number/K7DCIYGEQ2LWEJ22Q65EKL4U5J/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/K7DCIYGEQ2LWEJ22Q65EKL4U5J/action/timestamp_anchor","attest_storage":"https://pith.science/pith/K7DCIYGEQ2LWEJ22Q65EKL4U5J/action/storage_attestation","attest_author":"https://pith.science/pith/K7DCIYGEQ2LWEJ22Q65EKL4U5J/action/author_attestation","sign_citation":"https://pith.science/pith/K7DCIYGEQ2LWEJ22Q65EKL4U5J/action/citation_signature","submit_replication":"https://pith.science/pith/K7DCIYGEQ2LWEJ22Q65EKL4U5J/action/replication_record"}},"created_at":"2026-05-27T01:05:18.252093+00:00","updated_at":"2026-05-27T01:05:18.252093+00:00"}