{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:7LA3PIOHUCILP7YFZJAENBEFAQ","short_pith_number":"pith:7LA3PIOH","schema_version":"1.0","canonical_sha256":"fac1b7a1c7a090b7ff05ca404684850420b2228efc60b4c8ae911cef38ce23b3","source":{"kind":"arxiv","id":"2605.26172","version":1},"attestation_state":"computed","paper":{"title":"ARBITER: Reasoning Trajectory Basins and Majority Vote Failures in Test-Time Sampling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Farhana Choudhury, Lars Kulik, Meng Cai","submitted_at":"2026-05-25T04:07:44Z","abstract_excerpt":"When language models use test-time sampling, they generate multiple reasoning trajectories and select an answer by majority vote. We show that these trajectories are not independent: for a given question, they concentrate into a small number of clusters, or reasoning basins, each defined by a normalized final answer and the solutions that reach it. A majority vote therefore selects the most stable basin rather than the most accurate one, which creates wrong-majority failures where the correct answer is present but outvoted. We introduce ARBITER, a model-agnostic approach that models interactio"},"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":"2605.26172","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-25T04:07:44Z","cross_cats_sorted":[],"title_canon_sha256":"081fcf6672785e2e5ed84e637c9c9895cd1d22baa797490e0adef63821212f8a","abstract_canon_sha256":"9354e87c57c0e2285a74c7e44fb620806771483f427793fd8c38e438ec47dad7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-27T00:04:27.264027Z","signature_b64":"MrBHxD+eI17vUxcOxW3SWkYyTZE0hxX321poY/iXhCbosvGdLopx9CjAm9jqS1as0Y1wmLgaq3uJyM3qYbVTAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fac1b7a1c7a090b7ff05ca404684850420b2228efc60b4c8ae911cef38ce23b3","last_reissued_at":"2026-05-27T00:04:27.263218Z","signature_status":"signed_v1","first_computed_at":"2026-05-27T00:04:27.263218Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ARBITER: Reasoning Trajectory Basins and Majority Vote Failures in Test-Time Sampling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Farhana Choudhury, Lars Kulik, Meng Cai","submitted_at":"2026-05-25T04:07:44Z","abstract_excerpt":"When language models use test-time sampling, they generate multiple reasoning trajectories and select an answer by majority vote. We show that these trajectories are not independent: for a given question, they concentrate into a small number of clusters, or reasoning basins, each defined by a normalized final answer and the solutions that reach it. A majority vote therefore selects the most stable basin rather than the most accurate one, which creates wrong-majority failures where the correct answer is present but outvoted. We introduce ARBITER, a model-agnostic approach that models interactio"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.26172","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/2605.26172/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":"2605.26172","created_at":"2026-05-27T00:04:27.263357+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.26172v1","created_at":"2026-05-27T00:04:27.263357+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.26172","created_at":"2026-05-27T00:04:27.263357+00:00"},{"alias_kind":"pith_short_12","alias_value":"7LA3PIOHUCIL","created_at":"2026-05-27T00:04:27.263357+00:00"},{"alias_kind":"pith_short_16","alias_value":"7LA3PIOHUCILP7YF","created_at":"2026-05-27T00:04:27.263357+00:00"},{"alias_kind":"pith_short_8","alias_value":"7LA3PIOH","created_at":"2026-05-27T00:04:27.263357+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/7LA3PIOHUCILP7YFZJAENBEFAQ","json":"https://pith.science/pith/7LA3PIOHUCILP7YFZJAENBEFAQ.json","graph_json":"https://pith.science/api/pith-number/7LA3PIOHUCILP7YFZJAENBEFAQ/graph.json","events_json":"https://pith.science/api/pith-number/7LA3PIOHUCILP7YFZJAENBEFAQ/events.json","paper":"https://pith.science/paper/7LA3PIOH"},"agent_actions":{"view_html":"https://pith.science/pith/7LA3PIOHUCILP7YFZJAENBEFAQ","download_json":"https://pith.science/pith/7LA3PIOHUCILP7YFZJAENBEFAQ.json","view_paper":"https://pith.science/paper/7LA3PIOH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.26172&json=true","fetch_graph":"https://pith.science/api/pith-number/7LA3PIOHUCILP7YFZJAENBEFAQ/graph.json","fetch_events":"https://pith.science/api/pith-number/7LA3PIOHUCILP7YFZJAENBEFAQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7LA3PIOHUCILP7YFZJAENBEFAQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7LA3PIOHUCILP7YFZJAENBEFAQ/action/storage_attestation","attest_author":"https://pith.science/pith/7LA3PIOHUCILP7YFZJAENBEFAQ/action/author_attestation","sign_citation":"https://pith.science/pith/7LA3PIOHUCILP7YFZJAENBEFAQ/action/citation_signature","submit_replication":"https://pith.science/pith/7LA3PIOHUCILP7YFZJAENBEFAQ/action/replication_record"}},"created_at":"2026-05-27T00:04:27.263357+00:00","updated_at":"2026-05-27T00:04:27.263357+00:00"}