{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:FHK2SA2WL3NSCBN3HO2QLTJWUB","short_pith_number":"pith:FHK2SA2W","schema_version":"1.0","canonical_sha256":"29d5a903565edb2105bb3bb505cd36a06ff5d2690d62f52bcb5b9f5cdcb17098","source":{"kind":"arxiv","id":"2505.15107","version":2},"attestation_state":"computed","paper":{"title":"StepSearch: Igniting LLMs Search Ability via Step-Wise Proximal Policy Optimization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.IR"],"primary_cat":"cs.CL","authors_text":"Cijun Ouyang, Jialu Cai, Kang An, Xuhui Zheng, Yichao Wu, Yuhang Wang, Ziliang Wang","submitted_at":"2025-05-21T05:01:31Z","abstract_excerpt":"Efficient multi-hop reasoning requires Large Language Models (LLMs) based agents to acquire high-value external knowledge iteratively. Previous work has explored reinforcement learning (RL) to train LLMs to perform search-based document retrieval, achieving notable improvements in QA performance, but underperform on complex, multi-hop QA resulting from the sparse rewards from global signal only. To address this gap in existing research, we introduce StepSearch, a framework for search LLMs that trained with step-wise proximal policy optimization method. It consists of richer and more detailed i"},"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":"2505.15107","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-05-21T05:01:31Z","cross_cats_sorted":["cs.AI","cs.IR"],"title_canon_sha256":"d977de914fec6043096b22434bdc2c9db0be522f32507455d9040f7719f0fa1e","abstract_canon_sha256":"1106e155db072c9f1615361e92c37976eefc2ce8dc71326ea051d390cf320109"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:09:14.529783Z","signature_b64":"e2S3g7jjQbSIqb+kpDrj638srKXO6sRX3+DGlX3dJ1ApJMOy14TeSPE9/9jafiMQB/cbon+n/7LRQGdm68VCDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"29d5a903565edb2105bb3bb505cd36a06ff5d2690d62f52bcb5b9f5cdcb17098","last_reissued_at":"2026-07-05T11:09:14.529272Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:09:14.529272Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"StepSearch: Igniting LLMs Search Ability via Step-Wise Proximal Policy Optimization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.IR"],"primary_cat":"cs.CL","authors_text":"Cijun Ouyang, Jialu Cai, Kang An, Xuhui Zheng, Yichao Wu, Yuhang Wang, Ziliang Wang","submitted_at":"2025-05-21T05:01:31Z","abstract_excerpt":"Efficient multi-hop reasoning requires Large Language Models (LLMs) based agents to acquire high-value external knowledge iteratively. Previous work has explored reinforcement learning (RL) to train LLMs to perform search-based document retrieval, achieving notable improvements in QA performance, but underperform on complex, multi-hop QA resulting from the sparse rewards from global signal only. To address this gap in existing research, we introduce StepSearch, a framework for search LLMs that trained with step-wise proximal policy optimization method. It consists of richer and more detailed i"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.15107","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/2505.15107/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":"2505.15107","created_at":"2026-07-05T11:09:14.529346+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.15107v2","created_at":"2026-07-05T11:09:14.529346+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.15107","created_at":"2026-07-05T11:09:14.529346+00:00"},{"alias_kind":"pith_short_12","alias_value":"FHK2SA2WL3NS","created_at":"2026-07-05T11:09:14.529346+00:00"},{"alias_kind":"pith_short_16","alias_value":"FHK2SA2WL3NSCBN3","created_at":"2026-07-05T11:09:14.529346+00:00"},{"alias_kind":"pith_short_8","alias_value":"FHK2SA2W","created_at":"2026-07-05T11:09:14.529346+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":19,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.06223","citing_title":"Information Gain-based Rollout Policy Optimization: An Adaptive Tree-Structured Rollout Approach for Multi-Turn LLM Agents","ref_index":26,"is_internal_anchor":true},{"citing_arxiv_id":"2606.22995","citing_title":"Group-Graph Policy Optimization for Long-Horizon Agentic Reinforcement Learning","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2606.13316","citing_title":"ReSum: Synergizing LLM Reasoning and Summarization with Reinforcement Learning","ref_index":60,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09348","citing_title":"PBSD: Privileged Bayesian Self-Distillation for Long-Horizon Credit Assignment","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2606.08755","citing_title":"Co-Evolving Skill Generation and Policy Optimization","ref_index":94,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05885","citing_title":"When Denser Credit Is Not Enough: Evidence-Calibrated Policy Optimization for Long-Horizon LLM Agent Training","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02373","citing_title":"Harness-1: Reinforcement Learning for Search Agents with State-Externalizing Harnesses","ref_index":110,"is_internal_anchor":false},{"citing_arxiv_id":"2606.01249","citing_title":"Trust Region On-Policy Distillation","ref_index":119,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22511","citing_title":"Search-E1: Self-Distillation Drives Self-Evolution in Search-Augmented Reasoning","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22511","citing_title":"Search-E1: Self-Distillation Drives Self-Evolution in Search-Augmented Reasoning","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17734","citing_title":"Harnessing LLM Agents with Skill Programs","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18299","citing_title":"SD-Search: On-Policy Hindsight Self-Distillation for Search-Augmented Reasoning","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2509.02547","citing_title":"The Landscape of Agentic Reinforcement Learning for LLMs: A Survey","ref_index":280,"is_internal_anchor":false},{"citing_arxiv_id":"2510.00861","citing_title":"Erase to Improve: Erasable Reinforcement Learning for Search-Augmented LLMs","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09287","citing_title":"PiCA: Pivot-Based Credit Assignment for Search Agentic Reinforcement Learning","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09287","citing_title":"PiCA: Pivot-Based Credit Assignment for Search Agentic Reinforcement Learning","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2505.10978","citing_title":"Group-in-Group Policy Optimization for LLM Agent Training","ref_index":69,"is_internal_anchor":false},{"citing_arxiv_id":"2604.09455","citing_title":"E3-TIR: Enhanced Experience Exploitation for Tool-Integrated Reasoning","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14054","citing_title":"$\\pi$-Play: Multi-Agent Self-Play via Privileged Self-Distillation without External Data","ref_index":36,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FHK2SA2WL3NSCBN3HO2QLTJWUB","json":"https://pith.science/pith/FHK2SA2WL3NSCBN3HO2QLTJWUB.json","graph_json":"https://pith.science/api/pith-number/FHK2SA2WL3NSCBN3HO2QLTJWUB/graph.json","events_json":"https://pith.science/api/pith-number/FHK2SA2WL3NSCBN3HO2QLTJWUB/events.json","paper":"https://pith.science/paper/FHK2SA2W"},"agent_actions":{"view_html":"https://pith.science/pith/FHK2SA2WL3NSCBN3HO2QLTJWUB","download_json":"https://pith.science/pith/FHK2SA2WL3NSCBN3HO2QLTJWUB.json","view_paper":"https://pith.science/paper/FHK2SA2W","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.15107&json=true","fetch_graph":"https://pith.science/api/pith-number/FHK2SA2WL3NSCBN3HO2QLTJWUB/graph.json","fetch_events":"https://pith.science/api/pith-number/FHK2SA2WL3NSCBN3HO2QLTJWUB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FHK2SA2WL3NSCBN3HO2QLTJWUB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FHK2SA2WL3NSCBN3HO2QLTJWUB/action/storage_attestation","attest_author":"https://pith.science/pith/FHK2SA2WL3NSCBN3HO2QLTJWUB/action/author_attestation","sign_citation":"https://pith.science/pith/FHK2SA2WL3NSCBN3HO2QLTJWUB/action/citation_signature","submit_replication":"https://pith.science/pith/FHK2SA2WL3NSCBN3HO2QLTJWUB/action/replication_record"}},"created_at":"2026-07-05T11:09:14.529346+00:00","updated_at":"2026-07-05T11:09:14.529346+00:00"}