{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5UX7KIXH6WMWOVHZQL44663UVZ","short_pith_number":"pith:5UX7KIXH","schema_version":"1.0","canonical_sha256":"ed2ff522e7f5996754f982f9cf7b74ae76cc89a1268b464893723a4a01637887","source":{"kind":"arxiv","id":"2509.23730","version":2},"attestation_state":"computed","paper":{"title":"EAPO: Enhancing Policy Optimization with On-Demand Expert Assistance","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Cong Ma, Huaixiao Tou, Kai Jia, Minghao Li, Shiye Lei, Siyao Song, Ying Zeng, Zhihao Cheng","submitted_at":"2025-09-28T08:20:22Z","abstract_excerpt":"Large language models (LLMs) have recently advanced in reasoning when optimized with reinforcement learning (RL) under verifiable rewards. Existing methods primarily rely on outcome-based supervision to strengthen internal LLM reasoning, often leading to inefficient exploration and sparse rewards. To mitigate this issue, we propose Expert-Assisted Policy Optimization (EAPO), a novel RL framework that enhances exploration by incorporating multi-turn interactions with external experts during training. Unlike prior methods, where policies reason in isolation, EAPO incentivizes the policy to adapt"},"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":"2509.23730","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2025-09-28T08:20:22Z","cross_cats_sorted":[],"title_canon_sha256":"35a1100591d2e87ae803b35c3cddf54f9ad0277d5045b639159fe73298985a43","abstract_canon_sha256":"82f9fc13f486fd7560461c14718a49fafc93245f57c70fe204244655db11abca"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-29T02:05:36.522978Z","signature_b64":"8EqIU7ePw/mDLFbIpznIE5ob5xf89kxcXOYHUhnAhQSjgSK73EHO9IWa1mFwWqZpE7blWDa+EqKT7kyj7zEGBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ed2ff522e7f5996754f982f9cf7b74ae76cc89a1268b464893723a4a01637887","last_reissued_at":"2026-05-29T02:05:36.522538Z","signature_status":"signed_v1","first_computed_at":"2026-05-29T02:05:36.522538Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"EAPO: Enhancing Policy Optimization with On-Demand Expert Assistance","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Cong Ma, Huaixiao Tou, Kai Jia, Minghao Li, Shiye Lei, Siyao Song, Ying Zeng, Zhihao Cheng","submitted_at":"2025-09-28T08:20:22Z","abstract_excerpt":"Large language models (LLMs) have recently advanced in reasoning when optimized with reinforcement learning (RL) under verifiable rewards. Existing methods primarily rely on outcome-based supervision to strengthen internal LLM reasoning, often leading to inefficient exploration and sparse rewards. To mitigate this issue, we propose Expert-Assisted Policy Optimization (EAPO), a novel RL framework that enhances exploration by incorporating multi-turn interactions with external experts during training. Unlike prior methods, where policies reason in isolation, EAPO incentivizes the policy to adapt"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.23730","kind":"arxiv","version":2},"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/2509.23730/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":"2509.23730","created_at":"2026-05-29T02:05:36.522595+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.23730v2","created_at":"2026-05-29T02:05:36.522595+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.23730","created_at":"2026-05-29T02:05:36.522595+00:00"},{"alias_kind":"pith_short_12","alias_value":"5UX7KIXH6WMW","created_at":"2026-05-29T02:05:36.522595+00:00"},{"alias_kind":"pith_short_16","alias_value":"5UX7KIXH6WMWOVHZ","created_at":"2026-05-29T02:05:36.522595+00:00"},{"alias_kind":"pith_short_8","alias_value":"5UX7KIXH","created_at":"2026-05-29T02:05:36.522595+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/5UX7KIXH6WMWOVHZQL44663UVZ","json":"https://pith.science/pith/5UX7KIXH6WMWOVHZQL44663UVZ.json","graph_json":"https://pith.science/api/pith-number/5UX7KIXH6WMWOVHZQL44663UVZ/graph.json","events_json":"https://pith.science/api/pith-number/5UX7KIXH6WMWOVHZQL44663UVZ/events.json","paper":"https://pith.science/paper/5UX7KIXH"},"agent_actions":{"view_html":"https://pith.science/pith/5UX7KIXH6WMWOVHZQL44663UVZ","download_json":"https://pith.science/pith/5UX7KIXH6WMWOVHZQL44663UVZ.json","view_paper":"https://pith.science/paper/5UX7KIXH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.23730&json=true","fetch_graph":"https://pith.science/api/pith-number/5UX7KIXH6WMWOVHZQL44663UVZ/graph.json","fetch_events":"https://pith.science/api/pith-number/5UX7KIXH6WMWOVHZQL44663UVZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5UX7KIXH6WMWOVHZQL44663UVZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5UX7KIXH6WMWOVHZQL44663UVZ/action/storage_attestation","attest_author":"https://pith.science/pith/5UX7KIXH6WMWOVHZQL44663UVZ/action/author_attestation","sign_citation":"https://pith.science/pith/5UX7KIXH6WMWOVHZQL44663UVZ/action/citation_signature","submit_replication":"https://pith.science/pith/5UX7KIXH6WMWOVHZQL44663UVZ/action/replication_record"}},"created_at":"2026-05-29T02:05:36.522595+00:00","updated_at":"2026-05-29T02:05:36.522595+00:00"}