{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:KQ3KH2EUT2DCCAHZ6FTOZNVZI2","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"5a6ac865085b2664dc78e29bc353e509efef8d2704b7e734cf0b98455ba9cff6","cross_cats_sorted":["cs.CL","cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2025-04-04T04:41:28Z","title_canon_sha256":"7996dd2f6a35c010b6913abfcca9e480138c0124154c22fbbe4a9e4b99e57ace"},"schema_version":"1.0","source":{"id":"2504.03160","kind":"arxiv","version":4}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2504.03160","created_at":"2026-05-17T23:38:46Z"},{"alias_kind":"arxiv_version","alias_value":"2504.03160v4","created_at":"2026-05-17T23:38:46Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.03160","created_at":"2026-05-17T23:38:46Z"},{"alias_kind":"pith_short_12","alias_value":"KQ3KH2EUT2DC","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"KQ3KH2EUT2DCCAHZ","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"KQ3KH2EU","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:5a63d5d76efcd5294f930968118f68f31ed340883395fcf25f2d66d5f090f49a","target":"graph","created_at":"2026-05-17T23:38:46Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"DeepResearcher achieves substantial improvements of up to 28.9 points over prompt engineering-based baselines and up to 7.2 points over RAG-based RL agents, with emergent cognitive behaviors including planning, cross-validation, self-reflection, and honesty."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the multi-agent browsing architecture can reliably extract information from arbitrary real-world webpage structures at scale without introducing systematic biases or instability that would undermine the reported performance gains."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"End-to-end RL in authentic web environments produces LLM research agents that outperform prompt-engineering and RAG-based baselines by up to 28.9 and 7.2 points respectively while exhibiting emergent planning, cross-validation, and self-reflection."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"End-to-end RL training on the open web lets LLM agents outperform prompt and RAG baselines by up to 28.9 points while developing planning and self-reflection."}],"snapshot_sha256":"eea7bbcef4bd1c25a2bd03aed38de86b097ddc4e7746d2e24b0d8f87cf649252"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"Large Language Models (LLMs) equipped with web search capabilities have demonstrated impressive potential for deep research tasks. However, current approaches predominantly rely on either manually engineered prompts (prompt engineering-based) with brittle performance or reinforcement learning within controlled Retrieval-Augmented Generation (RAG) environments (RAG-based) that fail to capture the complexities of real-world interaction. In this paper, we introduce DeepResearcher, the first comprehensive framework for end-to-end training of LLM-based deep research agents through scaling reinforce","authors_text":"Dayuan Fu, Lyumanshan Ye, Pengfei Liu, Pengrui Lu, Xiangkun Hu, Xiaojie Cai, Yuxiang Zheng","cross_cats":["cs.CL","cs.LG"],"headline":"End-to-end RL training on the open web lets LLM agents outperform prompt and RAG baselines by up to 28.9 points while developing planning and self-reflection.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2025-04-04T04:41:28Z","title":"DeepResearcher: Scaling Deep Research via Reinforcement Learning in Real-world Environments"},"references":{"count":18,"internal_anchors":2,"resolved_work":18,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Qwen, :, An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi ","work_id":"e84e2134-b769-4027-8225-600d73ff84a9","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Timo Schick, Jane Dwivedi-Yu, Roberto Dess`ı, Roberta Raileanu, Maria Lomeli, Eric Hambro, Luke Zettle- moyer, Nicola Cancedda, and Thomas Scialom. 2023. Toolformer: Language models can teach themselv","work_id":"6fd18ea3-e439-4393-b21e-7dbddf8d3dd4","year":2023},{"cited_arxiv_id":"2503.05592","doi":"","is_internal_anchor":true,"ref_index":3,"title":"R1-Searcher: Incentivizing the Search Capability in LLMs via Reinforcement Learning","work_id":"f5ed73d2-f2ff-4cf6-853d-3586333e44ef","year":2025},{"cited_arxiv_id":"2501.12599","doi":"","is_internal_anchor":true,"ref_index":4,"title":"Kimi k1.5: Scaling Reinforcement Learning with LLMs","work_id":"bff96ab1-bd6a-4585-be23-74fdb51969c7","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, and Ashish Sabharwal. 2022. MuSiQue: Multihop questions via single-hop question composition. Transactions of the Association for Computational Lin","work_id":"9a1f5b9e-1481-4ef8-9826-dda294612579","year":2022}],"snapshot_sha256":"7eb220dfde26546444931055c723153fc67b25c640face9c0c5a4787d50e602a"},"source":{"id":"2504.03160","kind":"arxiv","version":4},"verdict":{"created_at":"2026-05-16T19:55:39.109853Z","id":"e8c3be02-a679-4ccc-b3f5-bf3cc6a75514","model_set":{"reader":"grok-4.3"},"one_line_summary":"End-to-end RL in authentic web environments produces LLM research agents that outperform prompt-engineering and RAG-based baselines by up to 28.9 and 7.2 points respectively while exhibiting emergent planning, cross-validation, and self-reflection.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"End-to-end RL training on the open web lets LLM agents outperform prompt and RAG baselines by up to 28.9 points while developing planning and self-reflection.","strongest_claim":"DeepResearcher achieves substantial improvements of up to 28.9 points over prompt engineering-based baselines and up to 7.2 points over RAG-based RL agents, with emergent cognitive behaviors including planning, cross-validation, self-reflection, and honesty.","weakest_assumption":"That the multi-agent browsing architecture can reliably extract information from arbitrary real-world webpage structures at scale without introducing systematic biases or instability that would undermine the reported performance gains."}},"verdict_id":"e8c3be02-a679-4ccc-b3f5-bf3cc6a75514"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:2a5b31c0fc93ff7a16c06e7febf31526c6723e77746fa4ba87f0636cffbdfcb0","target":"record","created_at":"2026-05-17T23:38:46Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"5a6ac865085b2664dc78e29bc353e509efef8d2704b7e734cf0b98455ba9cff6","cross_cats_sorted":["cs.CL","cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2025-04-04T04:41:28Z","title_canon_sha256":"7996dd2f6a35c010b6913abfcca9e480138c0124154c22fbbe4a9e4b99e57ace"},"schema_version":"1.0","source":{"id":"2504.03160","kind":"arxiv","version":4}},"canonical_sha256":"5436a3e8949e862100f9f166ecb6b94699bef612eebd8a025c2452e9a6a41bd3","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"5436a3e8949e862100f9f166ecb6b94699bef612eebd8a025c2452e9a6a41bd3","first_computed_at":"2026-05-17T23:38:46.762485Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:38:46.762485Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"lv8xdYESiWsZZXJtTa1rt7L3PS/5sdamd2Mr2kWXZTywODKu62bgEpAI78CB8ydAkcUbM0AXcICXQ0RbsoMjAQ==","signature_status":"signed_v1","signed_at":"2026-05-17T23:38:46.763010Z","signed_message":"canonical_sha256_bytes"},"source_id":"2504.03160","source_kind":"arxiv","source_version":4}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:2a5b31c0fc93ff7a16c06e7febf31526c6723e77746fa4ba87f0636cffbdfcb0","sha256:5a63d5d76efcd5294f930968118f68f31ed340883395fcf25f2d66d5f090f49a"],"state_sha256":"ab36d373c07ce3e0b884fc6ac69dafd8f5bef55553229f711375b2e7348cea4a"}