{"paper":{"title":"LiteResearcher: A Scalable Agentic RL Training Framework for Deep Research Agent","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A lite virtual world mirroring real searches lets a 4B agent master deep research via scalable RL.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Bince Qu, Bo Pan, Bo Zhang, Jianyu Zhang, Pan Zhang, Wanli Li, Wei Chen, Zheng Liu","submitted_at":"2026-04-20T08:11:09Z","abstract_excerpt":"Reinforcement Learning (RL) has emerged as a powerful training paradigm for LLM-based agents. However, scaling agentic RL for deep research remains constrained by two coupled challenges: hand-crafted synthetic data fails to elicit genuine real-world search capabilities, and real-world search dependency during RL training introduces instability and prohibitive cost, which limits the scalability of Agentic RL. LiteResearcher is a training framework that makes Agentic RL scalable: by constructing a lite virtual world that mirrors real-world search dynamics, we enable a continuously improving trai"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"LiteResearcher is a training framework that makes Agentic RL scalable: by constructing a lite virtual world that mirrors real-world search dynamics, we enable a continuously improving training recipe that empowers a tiny search agent to outperform large-scale open-source and commercial models (e.g., Tongyi DeepResearch and Claude-4.5 Sonnet). Specifically, on common benchmarks such as GAIA and Xbench, our LiteResearcher-4B achieves open-source state-of-the-art results of 71.3% and 78.0% respectively.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The lite virtual world accurately captures the essential dynamics of real-world search so that capabilities learned inside it transfer to genuine research tasks without introducing simulation-specific artifacts or instability.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"LiteResearcher uses a lite virtual world to make agentic RL training scalable and stable, enabling a 4B model to achieve 71.3% on GAIA and 78.0% on Xbench, outperforming larger open-source and commercial systems.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A lite virtual world mirroring real searches lets a 4B agent master deep research via scalable RL.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"3dcad5dc9b11949a3f06bc4197da4aacc986fe1e742b31e10fed41217ef75527"},"source":{"id":"2604.17931","kind":"arxiv","version":3},"verdict":{"id":"01bc0000-bce8-4f17-ab84-7cf860515925","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T04:18:00.687817Z","strongest_claim":"LiteResearcher is a training framework that makes Agentic RL scalable: by constructing a lite virtual world that mirrors real-world search dynamics, we enable a continuously improving training recipe that empowers a tiny search agent to outperform large-scale open-source and commercial models (e.g., Tongyi DeepResearch and Claude-4.5 Sonnet). Specifically, on common benchmarks such as GAIA and Xbench, our LiteResearcher-4B achieves open-source state-of-the-art results of 71.3% and 78.0% respectively.","one_line_summary":"LiteResearcher uses a lite virtual world to make agentic RL training scalable and stable, enabling a 4B model to achieve 71.3% on GAIA and 78.0% on Xbench, outperforming larger open-source and commercial systems.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The lite virtual world accurately captures the essential dynamics of real-world search so that capabilities learned inside it transfer to genuine research tasks without introducing simulation-specific artifacts or instability.","pith_extraction_headline":"A lite virtual world mirroring real searches lets a 4B agent master deep research via scalable RL."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.17931/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-20T04:35:54.579286Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"bcd52858755c8ccc6bc12aac2e5e6baefaca77b9c8ed724e5ea7e223ec4b80ef"},"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"}