{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:PQH3AW5BGZID7H6OVIFSX2BAJC","short_pith_number":"pith:PQH3AW5B","schema_version":"1.0","canonical_sha256":"7c0fb05ba136503f9fceaa0b2be820488d038f3222b7955ac6ce7480f25d7fb3","source":{"kind":"arxiv","id":"2608.03545","version":1},"attestation_state":"computed","paper":{"title":"Hi-TTRL: Regulating Consensus with Hints for Test-Time Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Kehai Chen, Kunbin Xu, Xingzuo Li, Xuefeng Bai","submitted_at":"2026-08-04T12:20:17Z","abstract_excerpt":"Test-time reinforcement learning (TTRL) improves the reasoning capabilities of large language models without labeled data by updating the policy with pseudo-labels constructed through majority voting. While effective, the reward signal assigned from majority voting is highly sensitive to consensus strength, defined as the frequency of the most common answer within a rollout group. In TTRL, consensus strength plays a dual role: it reflects both the reliability of the pseudo-label and the distribution of advantages. Low consensus can amplify updates from unreliable pseudo-labels through dispropo"},"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":"2608.03545","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-08-04T12:20:17Z","cross_cats_sorted":[],"title_canon_sha256":"adaf9b8b5e92fa6fe461a97b03fee23d78dc83c3ee36cb866ba226ce4b68fb2b","abstract_canon_sha256":"a6daeba3bc1fd3b0bb7f688795dd158e0f42fce5308390a57b9f2cc18cb186a0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-05T01:36:05.013243Z","signature_b64":"Cjrrkbsz4iO+D/9kJVmzn4KuszxfEPFee4CehMSiO/MNKOooNRMnQyKr9hhnMZfRpEWbnfKqXkirNesSg08XAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7c0fb05ba136503f9fceaa0b2be820488d038f3222b7955ac6ce7480f25d7fb3","last_reissued_at":"2026-08-05T01:36:05.011374Z","signature_status":"signed_v1","first_computed_at":"2026-08-05T01:36:05.011374Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Hi-TTRL: Regulating Consensus with Hints for Test-Time Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Kehai Chen, Kunbin Xu, Xingzuo Li, Xuefeng Bai","submitted_at":"2026-08-04T12:20:17Z","abstract_excerpt":"Test-time reinforcement learning (TTRL) improves the reasoning capabilities of large language models without labeled data by updating the policy with pseudo-labels constructed through majority voting. While effective, the reward signal assigned from majority voting is highly sensitive to consensus strength, defined as the frequency of the most common answer within a rollout group. In TTRL, consensus strength plays a dual role: it reflects both the reliability of the pseudo-label and the distribution of advantages. Low consensus can amplify updates from unreliable pseudo-labels through dispropo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.03545","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/2608.03545/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":"2608.03545","created_at":"2026-08-05T01:36:05.011991+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.03545v1","created_at":"2026-08-05T01:36:05.011991+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.03545","created_at":"2026-08-05T01:36:05.011991+00:00"},{"alias_kind":"pith_short_12","alias_value":"PQH3AW5BGZID","created_at":"2026-08-05T01:36:05.011991+00:00"},{"alias_kind":"pith_short_16","alias_value":"PQH3AW5BGZID7H6O","created_at":"2026-08-05T01:36:05.011991+00:00"},{"alias_kind":"pith_short_8","alias_value":"PQH3AW5B","created_at":"2026-08-05T01:36:05.011991+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/PQH3AW5BGZID7H6OVIFSX2BAJC","json":"https://pith.science/pith/PQH3AW5BGZID7H6OVIFSX2BAJC.json","graph_json":"https://pith.science/api/pith-number/PQH3AW5BGZID7H6OVIFSX2BAJC/graph.json","events_json":"https://pith.science/api/pith-number/PQH3AW5BGZID7H6OVIFSX2BAJC/events.json","paper":"https://pith.science/paper/PQH3AW5B"},"agent_actions":{"view_html":"https://pith.science/pith/PQH3AW5BGZID7H6OVIFSX2BAJC","download_json":"https://pith.science/pith/PQH3AW5BGZID7H6OVIFSX2BAJC.json","view_paper":"https://pith.science/paper/PQH3AW5B","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.03545&json=true","fetch_graph":"https://pith.science/api/pith-number/PQH3AW5BGZID7H6OVIFSX2BAJC/graph.json","fetch_events":"https://pith.science/api/pith-number/PQH3AW5BGZID7H6OVIFSX2BAJC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PQH3AW5BGZID7H6OVIFSX2BAJC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PQH3AW5BGZID7H6OVIFSX2BAJC/action/storage_attestation","attest_author":"https://pith.science/pith/PQH3AW5BGZID7H6OVIFSX2BAJC/action/author_attestation","sign_citation":"https://pith.science/pith/PQH3AW5BGZID7H6OVIFSX2BAJC/action/citation_signature","submit_replication":"https://pith.science/pith/PQH3AW5BGZID7H6OVIFSX2BAJC/action/replication_record"}},"created_at":"2026-08-05T01:36:05.011991+00:00","updated_at":"2026-08-05T01:36:05.011991+00:00"}