{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:VJBC57L2777PUCGJNBG3TGIZNG","short_pith_number":"pith:VJBC57L2","schema_version":"1.0","canonical_sha256":"aa422efd7afffefa08c9684db99919698a44700ec0f321a462322e594c2ec114","source":{"kind":"arxiv","id":"2508.13023","version":1},"attestation_state":"computed","paper":{"title":"G$^2$RPO-A: Guided Group Relative Policy Optimization with Adaptive Guidance","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Wenbo Deng, Xiaoying Tang, Yongxin Guo, Zhenglin Cheng","submitted_at":"2025-08-18T15:41:16Z","abstract_excerpt":"Reinforcement Learning with Verifiable Rewards (RLVR) has markedly enhanced the reasoning abilities of large language models (LLMs). Its success, however, largely depends on strong base models with rich world knowledge, yielding only modest improvements for small-size language models (SLMs). To address this limitation, we investigate Guided GRPO, which injects ground-truth reasoning steps into roll-out trajectories to compensate for SLMs' inherent weaknesses. Through a comprehensive study of various guidance configurations, we find that naively adding guidance delivers limited gains. These ins"},"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":"2508.13023","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.AI","submitted_at":"2025-08-18T15:41:16Z","cross_cats_sorted":[],"title_canon_sha256":"adf72d6642bc614907e410277597175e712a356a05aef3ae1025a095fa16929b","abstract_canon_sha256":"064eec4b7b4ac23d652d036fa46f7bb098442fc4f9190a1cdc16889060a67525"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:55:31.051200Z","signature_b64":"0d3DspdAf8uVejMVw36FFnybz48lLDCux8V9c27vD5OemHf5zXS/OoxeUunYRD8vOwEl5HmYS48BWg5wTxO5CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"aa422efd7afffefa08c9684db99919698a44700ec0f321a462322e594c2ec114","last_reissued_at":"2026-07-05T11:55:31.050781Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:55:31.050781Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"G$^2$RPO-A: Guided Group Relative Policy Optimization with Adaptive Guidance","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Wenbo Deng, Xiaoying Tang, Yongxin Guo, Zhenglin Cheng","submitted_at":"2025-08-18T15:41:16Z","abstract_excerpt":"Reinforcement Learning with Verifiable Rewards (RLVR) has markedly enhanced the reasoning abilities of large language models (LLMs). Its success, however, largely depends on strong base models with rich world knowledge, yielding only modest improvements for small-size language models (SLMs). To address this limitation, we investigate Guided GRPO, which injects ground-truth reasoning steps into roll-out trajectories to compensate for SLMs' inherent weaknesses. Through a comprehensive study of various guidance configurations, we find that naively adding guidance delivers limited gains. These ins"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.13023","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/2508.13023/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":"2508.13023","created_at":"2026-07-05T11:55:31.050837+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.13023v1","created_at":"2026-07-05T11:55:31.050837+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.13023","created_at":"2026-07-05T11:55:31.050837+00:00"},{"alias_kind":"pith_short_12","alias_value":"VJBC57L2777P","created_at":"2026-07-05T11:55:31.050837+00:00"},{"alias_kind":"pith_short_16","alias_value":"VJBC57L2777PUCGJ","created_at":"2026-07-05T11:55:31.050837+00:00"},{"alias_kind":"pith_short_8","alias_value":"VJBC57L2","created_at":"2026-07-05T11:55:31.050837+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.15980","citing_title":"Flash-GRPO: Efficient Alignment for Video Diffusion via One-Step Policy Optimization","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15980","citing_title":"Flash-GRPO: Efficient Alignment for Video Diffusion via One-Step Policy Optimization","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2509.08827","citing_title":"A Survey of Reinforcement Learning for Large Reasoning Models","ref_index":175,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02913","citing_title":"Generate, Filter, Control, Replay: A Comprehensive Survey of Rollout Strategies for LLM Reinforcement Learning","ref_index":30,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VJBC57L2777PUCGJNBG3TGIZNG","json":"https://pith.science/pith/VJBC57L2777PUCGJNBG3TGIZNG.json","graph_json":"https://pith.science/api/pith-number/VJBC57L2777PUCGJNBG3TGIZNG/graph.json","events_json":"https://pith.science/api/pith-number/VJBC57L2777PUCGJNBG3TGIZNG/events.json","paper":"https://pith.science/paper/VJBC57L2"},"agent_actions":{"view_html":"https://pith.science/pith/VJBC57L2777PUCGJNBG3TGIZNG","download_json":"https://pith.science/pith/VJBC57L2777PUCGJNBG3TGIZNG.json","view_paper":"https://pith.science/paper/VJBC57L2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.13023&json=true","fetch_graph":"https://pith.science/api/pith-number/VJBC57L2777PUCGJNBG3TGIZNG/graph.json","fetch_events":"https://pith.science/api/pith-number/VJBC57L2777PUCGJNBG3TGIZNG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VJBC57L2777PUCGJNBG3TGIZNG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VJBC57L2777PUCGJNBG3TGIZNG/action/storage_attestation","attest_author":"https://pith.science/pith/VJBC57L2777PUCGJNBG3TGIZNG/action/author_attestation","sign_citation":"https://pith.science/pith/VJBC57L2777PUCGJNBG3TGIZNG/action/citation_signature","submit_replication":"https://pith.science/pith/VJBC57L2777PUCGJNBG3TGIZNG/action/replication_record"}},"created_at":"2026-07-05T11:55:31.050837+00:00","updated_at":"2026-07-05T11:55:31.050837+00:00"}