{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:AETV3T3JX6ITSKDEDDJXC5UQBO","short_pith_number":"pith:AETV3T3J","schema_version":"1.0","canonical_sha256":"01275dcf69bf9139286418d37176900ba0349abe0e78e9dbb05d3b19019e74e0","source":{"kind":"arxiv","id":"2607.27770","version":1},"attestation_state":"computed","paper":{"title":"Beyond the Best Teacher: Expanding and Compressing the Reasoning Solution Manifold","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Songshuo Lu, Yaohua Tang, Zhi Chen","submitted_at":"2026-07-30T07:05:37Z","abstract_excerpt":"A single reinforcement-learning run can produce a strong reasoner yet an incomplete teacher: it often amplifies only a subset of the valid solution modes. We argue that reinforcement learning (RL)-trained policies should therefore be viewed as local probes of a multi-basin reasoning solution manifold, rather than as globally reliable supervisors. Based on this view, we propose an expand-then-compress framework that couples teacher construction with multi-teacher policy distillation. In the expansion stage, Residual Group Relative Policy Optimization (RGRPO) trains a sequence of teachers from a"},"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":"2607.27770","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-30T07:05:37Z","cross_cats_sorted":[],"title_canon_sha256":"047c8a7531b87dddb05226b30b241a0f30c0fde2a3ff8c906f8c87404af4afd7","abstract_canon_sha256":"d65125fa2d19a738a9ccd905e65ae1bccc97d9a69e0be934b6eafa201a72b25c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"01275dcf69bf9139286418d37176900ba0349abe0e78e9dbb05d3b19019e74e0","last_reissued_at":"2026-07-31T01:30:32.259318Z","signature_status":"unsigned_v0","first_computed_at":"2026-07-31T01:30:32.259318Z"},"graph_snapshot":{"paper":{"title":"Beyond the Best Teacher: Expanding and Compressing the Reasoning Solution Manifold","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Songshuo Lu, Yaohua Tang, Zhi Chen","submitted_at":"2026-07-30T07:05:37Z","abstract_excerpt":"A single reinforcement-learning run can produce a strong reasoner yet an incomplete teacher: it often amplifies only a subset of the valid solution modes. We argue that reinforcement learning (RL)-trained policies should therefore be viewed as local probes of a multi-basin reasoning solution manifold, rather than as globally reliable supervisors. Based on this view, we propose an expand-then-compress framework that couples teacher construction with multi-teacher policy distillation. In the expansion stage, Residual Group Relative Policy Optimization (RGRPO) trains a sequence of teachers from a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.27770","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/2607.27770/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":"2607.27770","created_at":"2026-07-31T01:30:32.262334+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.27770v1","created_at":"2026-07-31T01:30:32.262334+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.27770","created_at":"2026-07-31T01:30:32.262334+00:00"},{"alias_kind":"pith_short_12","alias_value":"AETV3T3JX6IT","created_at":"2026-07-31T01:30:32.262334+00:00"},{"alias_kind":"pith_short_16","alias_value":"AETV3T3JX6ITSKDE","created_at":"2026-07-31T01:30:32.262334+00:00"},{"alias_kind":"pith_short_8","alias_value":"AETV3T3J","created_at":"2026-07-31T01:30:32.262334+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/AETV3T3JX6ITSKDEDDJXC5UQBO","json":"https://pith.science/pith/AETV3T3JX6ITSKDEDDJXC5UQBO.json","graph_json":"https://pith.science/api/pith-number/AETV3T3JX6ITSKDEDDJXC5UQBO/graph.json","events_json":"https://pith.science/api/pith-number/AETV3T3JX6ITSKDEDDJXC5UQBO/events.json","paper":"https://pith.science/paper/AETV3T3J"},"agent_actions":{"view_html":"https://pith.science/pith/AETV3T3JX6ITSKDEDDJXC5UQBO","download_json":"https://pith.science/pith/AETV3T3JX6ITSKDEDDJXC5UQBO.json","view_paper":"https://pith.science/paper/AETV3T3J","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.27770&json=true","fetch_graph":"https://pith.science/api/pith-number/AETV3T3JX6ITSKDEDDJXC5UQBO/graph.json","fetch_events":"https://pith.science/api/pith-number/AETV3T3JX6ITSKDEDDJXC5UQBO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AETV3T3JX6ITSKDEDDJXC5UQBO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AETV3T3JX6ITSKDEDDJXC5UQBO/action/storage_attestation","attest_author":"https://pith.science/pith/AETV3T3JX6ITSKDEDDJXC5UQBO/action/author_attestation","sign_citation":"https://pith.science/pith/AETV3T3JX6ITSKDEDDJXC5UQBO/action/citation_signature","submit_replication":"https://pith.science/pith/AETV3T3JX6ITSKDEDDJXC5UQBO/action/replication_record"}},"created_at":"2026-07-31T01:30:32.262334+00:00","updated_at":"2026-07-31T01:30:32.262334+00:00"}