{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:WKOWY7ILQ3EHFH3KYXVCZPQ4MB","short_pith_number":"pith:WKOWY7IL","schema_version":"1.0","canonical_sha256":"b29d6c7d0b86c8729f6ac5ea2cbe1c60792fc8098a2181e76ceba8885367e1e6","source":{"kind":"arxiv","id":"2506.04708","version":3},"attestation_state":"computed","paper":{"title":"Accelerated Test-Time Scaling with Model-Free Speculative Sampling","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aram Galstyan, Bhavana Ganesh, Jinwoo Shin, Sai Muralidhar Jayanthi, Saket Dingliwal, Sravan Babu Bodapati, Woomin Song","submitted_at":"2025-06-05T07:31:18Z","abstract_excerpt":"Language models have demonstrated remarkable capabilities in reasoning tasks through test-time scaling techniques like best-of-N sampling and tree search. However, these approaches often demand substantial computational resources, creating a critical trade-off between performance and efficiency. We introduce STAND (STochastic Adaptive N-gram Drafting), a novel model-free speculative decoding approach that exploits the inherent redundancy in reasoning trajectories to achieve significant acceleration without compromising accuracy. Our analysis shows that reasoning paths frequently reuse similar "},"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":"2506.04708","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-05T07:31:18Z","cross_cats_sorted":[],"title_canon_sha256":"99e136240352e87a1aa0421ab25f434611604e5836fbe9e882dd308e2a69d62c","abstract_canon_sha256":"7f960c1e8104a74e50afe09aede4c244d80c5da398a797bc32643d4d48d9efaa"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T01:03:43.435399Z","signature_b64":"AdX8FMXert1LB+7poYi/RXyx1yurY9hJzgdf+vjAu4ucbfo27N0NSMmAcF4Cv85NcbRG+Z/BWdLN8N1KC/ruAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b29d6c7d0b86c8729f6ac5ea2cbe1c60792fc8098a2181e76ceba8885367e1e6","last_reissued_at":"2026-05-22T01:03:43.434646Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T01:03:43.434646Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Accelerated Test-Time Scaling with Model-Free Speculative Sampling","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aram Galstyan, Bhavana Ganesh, Jinwoo Shin, Sai Muralidhar Jayanthi, Saket Dingliwal, Sravan Babu Bodapati, Woomin Song","submitted_at":"2025-06-05T07:31:18Z","abstract_excerpt":"Language models have demonstrated remarkable capabilities in reasoning tasks through test-time scaling techniques like best-of-N sampling and tree search. However, these approaches often demand substantial computational resources, creating a critical trade-off between performance and efficiency. We introduce STAND (STochastic Adaptive N-gram Drafting), a novel model-free speculative decoding approach that exploits the inherent redundancy in reasoning trajectories to achieve significant acceleration without compromising accuracy. Our analysis shows that reasoning paths frequently reuse similar "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.04708","kind":"arxiv","version":3},"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/2506.04708/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":"2506.04708","created_at":"2026-05-22T01:03:43.434772+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.04708v3","created_at":"2026-05-22T01:03:43.434772+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.04708","created_at":"2026-05-22T01:03:43.434772+00:00"},{"alias_kind":"pith_short_12","alias_value":"WKOWY7ILQ3EH","created_at":"2026-05-22T01:03:43.434772+00:00"},{"alias_kind":"pith_short_16","alias_value":"WKOWY7ILQ3EHFH3K","created_at":"2026-05-22T01:03:43.434772+00:00"},{"alias_kind":"pith_short_8","alias_value":"WKOWY7IL","created_at":"2026-05-22T01:03:43.434772+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2503.16419","citing_title":"Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models","ref_index":161,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06165","citing_title":"Post Reasoning: Improving the Performance of Non-Thinking Models at No Cost","ref_index":227,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WKOWY7ILQ3EHFH3KYXVCZPQ4MB","json":"https://pith.science/pith/WKOWY7ILQ3EHFH3KYXVCZPQ4MB.json","graph_json":"https://pith.science/api/pith-number/WKOWY7ILQ3EHFH3KYXVCZPQ4MB/graph.json","events_json":"https://pith.science/api/pith-number/WKOWY7ILQ3EHFH3KYXVCZPQ4MB/events.json","paper":"https://pith.science/paper/WKOWY7IL"},"agent_actions":{"view_html":"https://pith.science/pith/WKOWY7ILQ3EHFH3KYXVCZPQ4MB","download_json":"https://pith.science/pith/WKOWY7ILQ3EHFH3KYXVCZPQ4MB.json","view_paper":"https://pith.science/paper/WKOWY7IL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.04708&json=true","fetch_graph":"https://pith.science/api/pith-number/WKOWY7ILQ3EHFH3KYXVCZPQ4MB/graph.json","fetch_events":"https://pith.science/api/pith-number/WKOWY7ILQ3EHFH3KYXVCZPQ4MB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WKOWY7ILQ3EHFH3KYXVCZPQ4MB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WKOWY7ILQ3EHFH3KYXVCZPQ4MB/action/storage_attestation","attest_author":"https://pith.science/pith/WKOWY7ILQ3EHFH3KYXVCZPQ4MB/action/author_attestation","sign_citation":"https://pith.science/pith/WKOWY7ILQ3EHFH3KYXVCZPQ4MB/action/citation_signature","submit_replication":"https://pith.science/pith/WKOWY7ILQ3EHFH3KYXVCZPQ4MB/action/replication_record"}},"created_at":"2026-05-22T01:03:43.434772+00:00","updated_at":"2026-05-22T01:03:43.434772+00:00"}