{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:YMYHA3CWK2DQOAODPCE7YEVEXU","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"cd7154bedad552d3beed7a7b2a5484f19224d573dfc42b98b25f01fbfdb32861","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-02-03T18:17:22Z","title_canon_sha256":"16717738a51e68b3ebc20dad8f8faae4f407ad3a18fc25924af456fae7e151cd"},"schema_version":"1.0","source":{"id":"2602.03814","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2602.03814","created_at":"2026-05-17T23:39:16Z"},{"alias_kind":"arxiv_version","alias_value":"2602.03814v2","created_at":"2026-05-17T23:39:16Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.03814","created_at":"2026-05-17T23:39:16Z"},{"alias_kind":"pith_short_12","alias_value":"YMYHA3CWK2DQ","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"YMYHA3CWK2DQOAOD","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"YMYHA3CW","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:334a616e7c1f218cf9b5560c74264a09cd90e0cc7a84da7b52a385d40015f450","target":"graph","created_at":"2026-05-17T23:39:16Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Given a target risk and a validation set, we use distribution-free risk control to optimally specify these stopping mechanisms. For scenarios with multiple budget controlling criteria, we incorporate an efficiency loss to select the most computationally efficient exiting mechanism."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The validation set is representative of future test instances so that the distribution-free risk guarantees transfer; the parametric form chosen for the lower threshold is flexible enough to capture unsolvable cases without excessive premature stopping."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Conformal risk control with upper and lower thresholds lets LLMs adaptively stop reasoning while guaranteeing a maximum error rate and minimizing token use."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Distribution-free risk control sets upper and lower thresholds so LLMs stop reasoning early while keeping error rates below a user target."}],"snapshot_sha256":"eb61cdec9e6ced04830e638a1f0c94d35e96bcfa5561e38448309840e575fb9e"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"487b269d6cc23194c947d055142fe5f8bd21392509f0a6156b82033ad5006ee0"},"paper":{"abstract_excerpt":"Reasoning Large Language Models (LLMs) enable test-time scaling, with dataset-level accuracy improving as the token budget increases, motivating adaptive reasoning -- spending tokens when they improve reliability and stopping early when additional computation is unlikely to help. However, setting the token budget, as well as the threshold for adaptive reasoning, is a practical challenge that entails a fundamental risk-accuracy trade-off. We re-frame the budget setting problem as risk control, limiting the error rate while minimizing compute. Our framework introduces an upper threshold that sto","authors_text":"Alvin Zhang, Anushri Suresh, Benjamin Van Durme, Daniel Khashabi, Eric Nalisnick, Mehrdad Farajtabar, Rishi More, William Jurayj, Xi Wang","cross_cats":["cs.LG"],"headline":"Distribution-free risk control sets upper and lower thresholds so LLMs stop reasoning early while keeping error rates below a user target.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-02-03T18:17:22Z","title":"Conformal Thinking: Risk Control for Reasoning on a Compute Budget"},"references":{"count":13,"internal_anchors":3,"resolved_work":13,"sample":[{"cited_arxiv_id":"2501.12948","doi":"","is_internal_anchor":true,"ref_index":1,"title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","work_id":"e6b75ad5-2877-4168-97c8-710407094d20","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Efficiently serving llm reasoning programs with certaindex","work_id":"57494d13-0d57-400a-8e38-67a61f1677d1","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"URL https: //aclanthology.org/2025.acl-short.50/. Langley, P. Crafting papers on machine learning. In Langley, P. (ed.),Proceedings of the 17th International Conference on Machine Learning (ICML 2000)","work_id":"1ec88e82-b25f-41d9-bd00-eb3bdab29850","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Fractured chain-of-thought reasoning","work_id":"bd0660ba-832d-46f0-90d4-ba4ae2efb956","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Xin Liu and Lu Wang","work_id":"125b44ab-fe4c-401f-9212-9a320bdf3efd","year":null}],"snapshot_sha256":"80749694133f90a9d50e2d789320e93f7e436ffca6d52d3ec737e0f953a24c4d"},"source":{"id":"2602.03814","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-16T07:43:16.978409Z","id":"5cbab2e9-02e3-4482-aab5-7ab1af81fa57","model_set":{"reader":"grok-4.3"},"one_line_summary":"Conformal risk control with upper and lower thresholds lets LLMs adaptively stop reasoning while guaranteeing a maximum error rate and minimizing token use.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Distribution-free risk control sets upper and lower thresholds so LLMs stop reasoning early while keeping error rates below a user target.","strongest_claim":"Given a target risk and a validation set, we use distribution-free risk control to optimally specify these stopping mechanisms. For scenarios with multiple budget controlling criteria, we incorporate an efficiency loss to select the most computationally efficient exiting mechanism.","weakest_assumption":"The validation set is representative of future test instances so that the distribution-free risk guarantees transfer; the parametric form chosen for the lower threshold is flexible enough to capture unsolvable cases without excessive premature stopping."}},"verdict_id":"5cbab2e9-02e3-4482-aab5-7ab1af81fa57"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:c7537fe5215efce06663fdd4fc84a5ccf41142c1c3b3be75eeaa06ff529b5f58","target":"record","created_at":"2026-05-17T23:39:16Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"cd7154bedad552d3beed7a7b2a5484f19224d573dfc42b98b25f01fbfdb32861","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-02-03T18:17:22Z","title_canon_sha256":"16717738a51e68b3ebc20dad8f8faae4f407ad3a18fc25924af456fae7e151cd"},"schema_version":"1.0","source":{"id":"2602.03814","kind":"arxiv","version":2}},"canonical_sha256":"c330706c5656870701c37889fc12a4bd0868230754bbb8c914059045df08a62f","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"c330706c5656870701c37889fc12a4bd0868230754bbb8c914059045df08a62f","first_computed_at":"2026-05-17T23:39:16.387425Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:39:16.387425Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"zRFaMTX++d8TD1v2unT982CcPwFh1TDddPGj0JmzdmpoS1NbSRap+kz28kXtNB470+VPX1iGtKWPP+/Q3U8oAA==","signature_status":"signed_v1","signed_at":"2026-05-17T23:39:16.388110Z","signed_message":"canonical_sha256_bytes"},"source_id":"2602.03814","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:c7537fe5215efce06663fdd4fc84a5ccf41142c1c3b3be75eeaa06ff529b5f58","sha256:334a616e7c1f218cf9b5560c74264a09cd90e0cc7a84da7b52a385d40015f450"],"state_sha256":"3c1df4d11e136491d174fd28a6bc71f2339aa62574897436f629b34c4000ee91"}