{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:7ZQRO7DFQKS7S7CQT4UL75RQAO","short_pith_number":"pith:7ZQRO7DF","schema_version":"1.0","canonical_sha256":"fe61177c6582a5f97c509f28bff6300384a1de7f51da37b74cd454c007285a6e","source":{"kind":"arxiv","id":"2606.12243","version":1},"attestation_state":"computed","paper":{"title":"VIA-SD: Verification via Intra-Model Routing for Speculative Decoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Yang He, Yi Yang, Yuchen Xian, Yunqiu Xu","submitted_at":"2026-06-10T15:45:18Z","abstract_excerpt":"Speculative decoding (SD) addresses the high inference costs of LLMs by having lightweight drafters generate candidates for large verifiers to validate in parallel. Existing draft-verify methods use binary decisions: accept or fully recompute. Yet we find that many rejected tokens can be verified correctly by a slim submodel derived from the full verifier via intra-model routing, instead of the full verifier. This motivates our slim-verifier to handle tokens requiring moderate verification resources, reducing expensive large-model calls. We propose Verification via Intra-Model Routing for Spec"},"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":"2606.12243","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-06-10T15:45:18Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"2e90cc81478873472ccc3cda252af33399c317908b051ca324224808f39970a1","abstract_canon_sha256":"d79a40057c37a11adb38d2e762e8a96e6691e9ec7ef98c55d82a7bde2d927e35"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-11T01:10:56.381782Z","signature_b64":"3AkR8xUDAysKZdICTD/L11Crr74aZEDU6t1HvyIRLtzDfS3BFm9YeH2idX9JJYcXYBhbHAzQJAaI2h7U9/ihDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fe61177c6582a5f97c509f28bff6300384a1de7f51da37b74cd454c007285a6e","last_reissued_at":"2026-06-11T01:10:56.381067Z","signature_status":"signed_v1","first_computed_at":"2026-06-11T01:10:56.381067Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"VIA-SD: Verification via Intra-Model Routing for Speculative Decoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Yang He, Yi Yang, Yuchen Xian, Yunqiu Xu","submitted_at":"2026-06-10T15:45:18Z","abstract_excerpt":"Speculative decoding (SD) addresses the high inference costs of LLMs by having lightweight drafters generate candidates for large verifiers to validate in parallel. Existing draft-verify methods use binary decisions: accept or fully recompute. Yet we find that many rejected tokens can be verified correctly by a slim submodel derived from the full verifier via intra-model routing, instead of the full verifier. This motivates our slim-verifier to handle tokens requiring moderate verification resources, reducing expensive large-model calls. We propose Verification via Intra-Model Routing for Spec"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.12243","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/2606.12243/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":"2606.12243","created_at":"2026-06-11T01:10:56.381191+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.12243v1","created_at":"2026-06-11T01:10:56.381191+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.12243","created_at":"2026-06-11T01:10:56.381191+00:00"},{"alias_kind":"pith_short_12","alias_value":"7ZQRO7DFQKS7","created_at":"2026-06-11T01:10:56.381191+00:00"},{"alias_kind":"pith_short_16","alias_value":"7ZQRO7DFQKS7S7CQ","created_at":"2026-06-11T01:10:56.381191+00:00"},{"alias_kind":"pith_short_8","alias_value":"7ZQRO7DF","created_at":"2026-06-11T01:10:56.381191+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/7ZQRO7DFQKS7S7CQT4UL75RQAO","json":"https://pith.science/pith/7ZQRO7DFQKS7S7CQT4UL75RQAO.json","graph_json":"https://pith.science/api/pith-number/7ZQRO7DFQKS7S7CQT4UL75RQAO/graph.json","events_json":"https://pith.science/api/pith-number/7ZQRO7DFQKS7S7CQT4UL75RQAO/events.json","paper":"https://pith.science/paper/7ZQRO7DF"},"agent_actions":{"view_html":"https://pith.science/pith/7ZQRO7DFQKS7S7CQT4UL75RQAO","download_json":"https://pith.science/pith/7ZQRO7DFQKS7S7CQT4UL75RQAO.json","view_paper":"https://pith.science/paper/7ZQRO7DF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.12243&json=true","fetch_graph":"https://pith.science/api/pith-number/7ZQRO7DFQKS7S7CQT4UL75RQAO/graph.json","fetch_events":"https://pith.science/api/pith-number/7ZQRO7DFQKS7S7CQT4UL75RQAO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7ZQRO7DFQKS7S7CQT4UL75RQAO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7ZQRO7DFQKS7S7CQT4UL75RQAO/action/storage_attestation","attest_author":"https://pith.science/pith/7ZQRO7DFQKS7S7CQT4UL75RQAO/action/author_attestation","sign_citation":"https://pith.science/pith/7ZQRO7DFQKS7S7CQT4UL75RQAO/action/citation_signature","submit_replication":"https://pith.science/pith/7ZQRO7DFQKS7S7CQT4UL75RQAO/action/replication_record"}},"created_at":"2026-06-11T01:10:56.381191+00:00","updated_at":"2026-06-11T01:10:56.381191+00:00"}