{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:FC5CRF5AJN7DLSKVMONNVP4367","short_pith_number":"pith:FC5CRF5A","schema_version":"1.0","canonical_sha256":"28ba2897a04b7e35c955639adabf9bf7c3ef5302f564bc691079c39e60c41bc8","source":{"kind":"arxiv","id":"2605.19444","version":1},"attestation_state":"computed","paper":{"title":"When the Majority Votes Wrong, the Intervention Timing for Test-Time Reinforcement Learning Hides in the Extinction Window","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Erpeng Xue, Hongxiang Lin, Lei Wang, Zhirui Kuai","submitted_at":"2026-05-19T06:58:44Z","abstract_excerpt":"Test-time reinforcement learning (TTRL) reports substantial accuracy gains on mathematical reasoning benchmarks using majority vote as a pseudo-label signal. We argue these gains are systematically misinterpreted: most reflect sharpening of already-solvable problems rather than genuine learning, while problems corrupted from correct to incorrect outnumber truly learned ones, and this damage is irreversible once majority vote locks onto a wrong answer. Per-problem tracking reveals that correct-answer signals in low-ability problems are briefly active before being permanently suppressed, a pheno"},"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":"2605.19444","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-19T06:58:44Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"f263114c952ffa03281ea5ac95af67c89704bc3bd72680d355245d601ec2ebae","abstract_canon_sha256":"827d52b187e09aad0b2a3e7851f17d29a4065129d62cf47df241802b8252f096"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T01:05:46.197920Z","signature_b64":"FbxyQyqkAquTFnJS0WJiDTlS+XEwr44/R49LqrAX6OLZELRirYTChh52/BmXfYIyOpjdoOLCEtQFH5GnfZUICg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"28ba2897a04b7e35c955639adabf9bf7c3ef5302f564bc691079c39e60c41bc8","last_reissued_at":"2026-05-20T01:05:46.196997Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T01:05:46.196997Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"When the Majority Votes Wrong, the Intervention Timing for Test-Time Reinforcement Learning Hides in the Extinction Window","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Erpeng Xue, Hongxiang Lin, Lei Wang, Zhirui Kuai","submitted_at":"2026-05-19T06:58:44Z","abstract_excerpt":"Test-time reinforcement learning (TTRL) reports substantial accuracy gains on mathematical reasoning benchmarks using majority vote as a pseudo-label signal. We argue these gains are systematically misinterpreted: most reflect sharpening of already-solvable problems rather than genuine learning, while problems corrupted from correct to incorrect outnumber truly learned ones, and this damage is irreversible once majority vote locks onto a wrong answer. Per-problem tracking reveals that correct-answer signals in low-ability problems are briefly active before being permanently suppressed, a pheno"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.19444","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/2605.19444/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":"2605.19444","created_at":"2026-05-20T01:05:46.197160+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.19444v1","created_at":"2026-05-20T01:05:46.197160+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.19444","created_at":"2026-05-20T01:05:46.197160+00:00"},{"alias_kind":"pith_short_12","alias_value":"FC5CRF5AJN7D","created_at":"2026-05-20T01:05:46.197160+00:00"},{"alias_kind":"pith_short_16","alias_value":"FC5CRF5AJN7DLSKV","created_at":"2026-05-20T01:05:46.197160+00:00"},{"alias_kind":"pith_short_8","alias_value":"FC5CRF5A","created_at":"2026-05-20T01:05:46.197160+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/FC5CRF5AJN7DLSKVMONNVP4367","json":"https://pith.science/pith/FC5CRF5AJN7DLSKVMONNVP4367.json","graph_json":"https://pith.science/api/pith-number/FC5CRF5AJN7DLSKVMONNVP4367/graph.json","events_json":"https://pith.science/api/pith-number/FC5CRF5AJN7DLSKVMONNVP4367/events.json","paper":"https://pith.science/paper/FC5CRF5A"},"agent_actions":{"view_html":"https://pith.science/pith/FC5CRF5AJN7DLSKVMONNVP4367","download_json":"https://pith.science/pith/FC5CRF5AJN7DLSKVMONNVP4367.json","view_paper":"https://pith.science/paper/FC5CRF5A","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.19444&json=true","fetch_graph":"https://pith.science/api/pith-number/FC5CRF5AJN7DLSKVMONNVP4367/graph.json","fetch_events":"https://pith.science/api/pith-number/FC5CRF5AJN7DLSKVMONNVP4367/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FC5CRF5AJN7DLSKVMONNVP4367/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FC5CRF5AJN7DLSKVMONNVP4367/action/storage_attestation","attest_author":"https://pith.science/pith/FC5CRF5AJN7DLSKVMONNVP4367/action/author_attestation","sign_citation":"https://pith.science/pith/FC5CRF5AJN7DLSKVMONNVP4367/action/citation_signature","submit_replication":"https://pith.science/pith/FC5CRF5AJN7DLSKVMONNVP4367/action/replication_record"}},"created_at":"2026-05-20T01:05:46.197160+00:00","updated_at":"2026-05-20T01:05:46.197160+00:00"}