{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:IK2E4IUYZFRU7LQF2Z6YTAY5HB","short_pith_number":"pith:IK2E4IUY","schema_version":"1.0","canonical_sha256":"42b44e2298c9634fae05d67d89831d38484b2af37706825e93ffef3896928511","source":{"kind":"arxiv","id":"2509.06980","version":1},"attestation_state":"computed","paper":{"title":"RLFactory: A Plug-and-Play Reinforcement Learning Post-Training Framework for LLM Multi-Turn Tool-Use","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Chengqi Dong, Chuhuai Yue, Guojun Yin, Hang He, Jiajun Chai, Jiwen Jiang, Siyu Xia, Wei Lin, Xiaoguang Li, Xiaohan Wang, Yi Jia, Zekun Xu","submitted_at":"2025-08-31T16:47:31Z","abstract_excerpt":"Large language models excel at basic reasoning but struggle with tasks that require interaction with external tools. We present RLFactory, a plug-and-play reinforcement learning post-training framework for multi-round tool use. RLFactory tackles (i) tool-call stability and adaptability amid tool heterogeneity and interface issues via an asyncio-based asynchronous caller and a decoupled tool/training architecture, and (ii) diverse evaluation needs via a reward layer supporting rule-based, model-judgment, and tool-verification signals. It reconstructs the MDP by introducing observation markers f"},"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":"2509.06980","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-08-31T16:47:31Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"fb6eb9e7fccb14f5af21b8c9da84492e1eb5450e864f72aaa4978d1c60bbfe4c","abstract_canon_sha256":"a53f921cc329ac2f2be841fb905d25ce8c27eb79bfc2e82ac4837222be2f50a9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:07:13.401456Z","signature_b64":"qWIv611HrPKqEW7WnArKychAE5B/mFAJ7teFRiDsshxOW7uvih05qVjuwSIc3/jKlojxh32DF0D/ZMD9Yc54AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"42b44e2298c9634fae05d67d89831d38484b2af37706825e93ffef3896928511","last_reissued_at":"2026-07-05T12:07:13.400928Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:07:13.400928Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"RLFactory: A Plug-and-Play Reinforcement Learning Post-Training Framework for LLM Multi-Turn Tool-Use","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Chengqi Dong, Chuhuai Yue, Guojun Yin, Hang He, Jiajun Chai, Jiwen Jiang, Siyu Xia, Wei Lin, Xiaoguang Li, Xiaohan Wang, Yi Jia, Zekun Xu","submitted_at":"2025-08-31T16:47:31Z","abstract_excerpt":"Large language models excel at basic reasoning but struggle with tasks that require interaction with external tools. We present RLFactory, a plug-and-play reinforcement learning post-training framework for multi-round tool use. RLFactory tackles (i) tool-call stability and adaptability amid tool heterogeneity and interface issues via an asyncio-based asynchronous caller and a decoupled tool/training architecture, and (ii) diverse evaluation needs via a reward layer supporting rule-based, model-judgment, and tool-verification signals. It reconstructs the MDP by introducing observation markers f"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.06980","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/2509.06980/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":"2509.06980","created_at":"2026-07-05T12:07:13.400996+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.06980v1","created_at":"2026-07-05T12:07:13.400996+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.06980","created_at":"2026-07-05T12:07:13.400996+00:00"},{"alias_kind":"pith_short_12","alias_value":"IK2E4IUYZFRU","created_at":"2026-07-05T12:07:13.400996+00:00"},{"alias_kind":"pith_short_16","alias_value":"IK2E4IUYZFRU7LQF","created_at":"2026-07-05T12:07:13.400996+00:00"},{"alias_kind":"pith_short_8","alias_value":"IK2E4IUY","created_at":"2026-07-05T12:07:13.400996+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.05784","citing_title":"TAPO: Tool-Aware Policy Optimization via Credit Transfer for Multimodal Search Agents","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03273","citing_title":"VistaHop: Benchmarking Long-Horizon Visual DeepSearch","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07791","citing_title":"SEARL: Joint Optimization of Policy and Tool Graph Memory for Self-Evolving Agents","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IK2E4IUYZFRU7LQF2Z6YTAY5HB","json":"https://pith.science/pith/IK2E4IUYZFRU7LQF2Z6YTAY5HB.json","graph_json":"https://pith.science/api/pith-number/IK2E4IUYZFRU7LQF2Z6YTAY5HB/graph.json","events_json":"https://pith.science/api/pith-number/IK2E4IUYZFRU7LQF2Z6YTAY5HB/events.json","paper":"https://pith.science/paper/IK2E4IUY"},"agent_actions":{"view_html":"https://pith.science/pith/IK2E4IUYZFRU7LQF2Z6YTAY5HB","download_json":"https://pith.science/pith/IK2E4IUYZFRU7LQF2Z6YTAY5HB.json","view_paper":"https://pith.science/paper/IK2E4IUY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.06980&json=true","fetch_graph":"https://pith.science/api/pith-number/IK2E4IUYZFRU7LQF2Z6YTAY5HB/graph.json","fetch_events":"https://pith.science/api/pith-number/IK2E4IUYZFRU7LQF2Z6YTAY5HB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IK2E4IUYZFRU7LQF2Z6YTAY5HB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IK2E4IUYZFRU7LQF2Z6YTAY5HB/action/storage_attestation","attest_author":"https://pith.science/pith/IK2E4IUYZFRU7LQF2Z6YTAY5HB/action/author_attestation","sign_citation":"https://pith.science/pith/IK2E4IUYZFRU7LQF2Z6YTAY5HB/action/citation_signature","submit_replication":"https://pith.science/pith/IK2E4IUYZFRU7LQF2Z6YTAY5HB/action/replication_record"}},"created_at":"2026-07-05T12:07:13.400996+00:00","updated_at":"2026-07-05T12:07:13.400996+00:00"}