https://github.com/plageon/HierSearch.

🌹 If you find this work helpful, please ✨star our GitHub repository or upvote this paper to support us. Your star means a lot!

\n","updatedAt":"2025-08-13T02:38:11.829Z","author":{"_id":"62e52483a944e2a56cd2c6ca","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/62e52483a944e2a56cd2c6ca/xElM_6teIrP3QI-Run0Bl.jpeg","fullname":"Jiejun Tan","name":"zstanjj","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":18,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.754859983921051},"editors":["zstanjj"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/62e52483a944e2a56cd2c6ca/xElM_6teIrP3QI-Run0Bl.jpeg"],"reactions":[{"reaction":"❤️","users":["dongguanting"],"count":1},{"reaction":"🔥","users":["dongguanting"],"count":1}],"isReport":false}},{"id":"689c3c14c2245c679b3f1f76","author":{"_id":"61cd4b833dd34ba1985e0753","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/61cd4b833dd34ba1985e0753/BfHfrwotoMESpXZOHiIe4.png","fullname":"KABI","name":"dongguanting","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":64,"isUserFollowing":false},"createdAt":"2025-08-13T07:17:40.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"An intriguing and practical RL design for deep search！","html":"

An intriguing and practical RL design for deep search！

\n","updatedAt":"2025-08-13T07:17:40.412Z","author":{"_id":"61cd4b833dd34ba1985e0753","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/61cd4b833dd34ba1985e0753/BfHfrwotoMESpXZOHiIe4.png","fullname":"KABI","name":"dongguanting","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":64,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.717340350151062},"editors":["dongguanting"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/61cd4b833dd34ba1985e0753/BfHfrwotoMESpXZOHiIe4.png"],"reactions":[{"reaction":"🤗","users":["zstanjj","SnowNation"],"count":2},{"reaction":"👀","users":["SnowNation"],"count":1},{"reaction":"🤝","users":["SnowNation"],"count":1}],"isReport":false}},{"id":"689d3e134b0d0be1118ae541","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":318,"isUserFollowing":false},"createdAt":"2025-08-14T01:38:27.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"This is an automated message from the [Librarian Bot](https://huggingface.co/librarian-bots). I found the following papers similar to this paper. \n\nThe following papers were recommended by the Semantic Scholar API \n\n* [Decoupled Planning and Execution: A Hierarchical Reasoning Framework for Deep Search](https://huggingface.co/papers/2507.02652) (2025)\n* [ReasonRank: Empowering Passage Ranking with Strong Reasoning Ability](https://huggingface.co/papers/2508.07050) (2025)\n* [WebWatcher: Breaking New Frontier of Vision-Language Deep Research Agent](https://huggingface.co/papers/2508.05748) (2025)\n* [BrowseMaster: Towards Scalable Web Browsing via Tool-Augmented Programmatic Agent Pair](https://huggingface.co/papers/2508.09129) (2025)\n* [Careful Queries, Credible Results: Teaching RAG Models Advanced Web Search Tools with Reinforcement Learning](https://huggingface.co/papers/2508.07956) (2025)\n* [RAG-R1 : Incentivize the Search and Reasoning Capabilities of LLMs through Multi-query Parallelism](https://huggingface.co/papers/2507.02962) (2025)\n* [DeepSieve: Information Sieving via LLM-as-a-Knowledge-Router](https://huggingface.co/papers/2507.22050) (2025)\n\n\n Please give a thumbs up to this comment if you found it helpful!\n\n If you want recommendations for any Paper on Hugging Face checkout [this](https://huggingface.co/spaces/librarian-bots/recommend_similar_papers) Space\n\n You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: `@librarian-bot recommend`","html":"

This is an automated message from the Librarian Bot. I found the following papers similar to this paper.

The following papers were recommended by the Semantic Scholar API

Please give a thumbs up to this comment if you found it helpful!

If you want recommendations for any Paper on Hugging Face checkout this Space

You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: \n\n@librarian-bot\n\t recommend

\n","updatedAt":"2025-08-14T01:38:27.574Z","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":318,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.6979691386222839},"editors":["librarian-bot"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg"],"reactions":[{"reaction":"👀","users":["zstanjj"],"count":1}],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2508.08088","authors":[{"_id":"689ac80efab6fdd2e52ac510","user":{"_id":"62e52483a944e2a56cd2c6ca","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/62e52483a944e2a56cd2c6ca/xElM_6teIrP3QI-Run0Bl.jpeg","isPro":false,"fullname":"Jiejun Tan","user":"zstanjj","type":"user"},"name":"Jiejun Tan","status":"claimed_verified","statusLastChangedAt":"2025-08-13T07:19:26.764Z","hidden":false},{"_id":"689ac80efab6fdd2e52ac511","user":{"_id":"66f0bf59e9d50ec57febf751","avatarUrl":"/avatars/be97941e60064e5dd806c6fe9db3c537.svg","isPro":false,"fullname":"Zhicheng Dou","user":"douzc","type":"user"},"name":"Zhicheng Dou","status":"claimed_verified","statusLastChangedAt":"2025-11-06T11:50:43.761Z","hidden":false},{"_id":"689ac80efab6fdd2e52ac512","name":"Yan Yu","hidden":false},{"_id":"689ac80efab6fdd2e52ac513","name":"Jiehan Cheng","hidden":false},{"_id":"689ac80efab6fdd2e52ac514","user":{"_id":"62dcdb86d36b2070f928a51e","avatarUrl":"/avatars/a341e4305217f8abd14cff97201a24aa.svg","isPro":false,"fullname":"sdujq","user":"sdujq","type":"user"},"name":"Qiang Ju","status":"claimed_verified","statusLastChangedAt":"2025-09-03T08:30:15.134Z","hidden":false},{"_id":"689ac80efab6fdd2e52ac515","name":"Jian Xie","hidden":false},{"_id":"689ac80efab6fdd2e52ac516","name":"Ji-Rong Wen","hidden":false}],"publishedAt":"2025-08-11T15:31:47.000Z","submittedOnDailyAt":"2025-08-13T01:08:11.771Z","title":"HierSearch: A Hierarchical Enterprise Deep Search Framework Integrating\n Local and Web Searches","submittedOnDailyBy":{"_id":"62e52483a944e2a56cd2c6ca","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/62e52483a944e2a56cd2c6ca/xElM_6teIrP3QI-Run0Bl.jpeg","isPro":false,"fullname":"Jiejun Tan","user":"zstanjj","type":"user"},"summary":"Recently, large reasoning models have demonstrated strong mathematical and\ncoding abilities, and deep search leverages their reasoning capabilities in\nchallenging information retrieval tasks. Existing deep search works are\ngenerally limited to a single knowledge source, either local or the Web.\nHowever, enterprises often require private deep search systems that can\nleverage search tools over both local and the Web corpus. Simply training an\nagent equipped with multiple search tools using flat reinforcement learning\n(RL) is a straightforward idea, but it has problems such as low training data\nefficiency and poor mastery of complex tools. To address the above issue, we\npropose a hierarchical agentic deep search framework, HierSearch, trained with\nhierarchical RL. At the low level, a local deep search agent and a Web deep\nsearch agent are trained to retrieve evidence from their corresponding domains.\nAt the high level, a planner agent coordinates low-level agents and provides\nthe final answer. Moreover, to prevent direct answer copying and error\npropagation, we design a knowledge refiner that filters out hallucinations and\nirrelevant evidence returned by low-level agents. Experiments show that\nHierSearch achieves better performance compared to flat RL, and outperforms\nvarious deep search and multi-source retrieval-augmented generation baselines\nin six benchmarks across general, finance, and medical domains.","upvotes":29,"discussionId":"689ac80efab6fdd2e52ac517","githubRepo":"https://github.com/plageon/HierSearch","githubRepoAddedBy":"user","ai_summary":"HierSearch, a hierarchical agentic deep search framework using hierarchical RL, improves performance in multi-source retrieval tasks by coordinating local and Web search agents and refining knowledge.","ai_keywords":["reinforcement learning","hierarchical RL","local deep search agent","Web deep search agent","planner agent","knowledge refiner","hallucinations","irrelevant evidence","multi-source retrieval-augmented generation","benchmarks"],"githubStars":37},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"62e52483a944e2a56cd2c6ca","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/62e52483a944e2a56cd2c6ca/xElM_6teIrP3QI-Run0Bl.jpeg","isPro":false,"fullname":"Jiejun Tan","user":"zstanjj","type":"user"},{"_id":"644a4fbc2166258fccc664bc","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/8k3b44MbhQiWuo6i8BnYl.jpeg","isPro":false,"fullname":"Jialong Wu","user":"callanwu","type":"user"},{"_id":"64d068a231c655ff8a77153e","avatarUrl":"/avatars/2b7407be92b65d435fecc3c29e7f8455.svg","isPro":false,"fullname":"wenhan liu","user":"liuwenhan","type":"user"},{"_id":"66b24ce4963f54b01727f4a8","avatarUrl":"/avatars/a490890fb288f5b7120d7b2f02e5482a.svg","isPro":false,"fullname":"The Licker","user":"CyberLick","type":"user"},{"_id":"682c0fcebbe0c6fa323f531b","avatarUrl":"/avatars/953e6f5ce0361c2f9693bc0ca82787b7.svg","isPro":false,"fullname":"yy","user":"yuy07","type":"user"},{"_id":"6695f14df0ffd8e3a379ad61","avatarUrl":"/avatars/5ebb7e55ee9c2d93850b279f440675b0.svg","isPro":false,"fullname":"Jiajie Jin","user":"jinjiajie","type":"user"},{"_id":"6722fe4a43b00f3c889aead4","avatarUrl":"/avatars/cbb57f517c37c72642b80dd538271238.svg","isPro":false,"fullname":"Xinyu Yang","user":"yxy919","type":"user"},{"_id":"62dcdb86d36b2070f928a51e","avatarUrl":"/avatars/a341e4305217f8abd14cff97201a24aa.svg","isPro":false,"fullname":"sdujq","user":"sdujq","type":"user"},{"_id":"66d81fe818e470c7312c0f2a","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/66d81fe818e470c7312c0f2a/HwPcM_yDB0nAgjz9bT_Cs.jpeg","isPro":false,"fullname":"ShutingWang","user":"ShootingWong","type":"user"},{"_id":"6544b9b646dbdeca34ee5f52","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6544b9b646dbdeca34ee5f52/nRx6m1C4wfZ_xSWoBUNJf.png","isPro":false,"fullname":"Yuyang Hu","user":"namespace-ERI","type":"user"},{"_id":"67a2db84756328af7d0b0ae7","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/67a2db84756328af7d0b0ae7/WlurtnmY59Mxe1B7VrTxu.jpeg","isPro":false,"fullname":"Jincheng Feng","user":"JinchengFeng","type":"user"},{"_id":"65db23d1f386d08eb0d1cec5","avatarUrl":"/avatars/b495ec5b35b15fea245ef490b83d1856.svg","isPro":false,"fullname":"Mengjie Deng","user":"MengjieDeng","type":"user"}],"acceptLanguages":["*"],"dailyPaperRank":0}">

Papers

arxiv:2508.08088

HierSearch: A Hierarchical Enterprise Deep Search Framework Integrating Local and Web Searches

Published on Aug 11, 2025

· Submitted by

Jiejun Tan on Aug 13, 2025

Upvote

Authors:

Jiejun Tan ,

Zhicheng Dou ,

Qiang Ju ,

Abstract

HierSearch, a hierarchical agentic deep search framework using hierarchical RL, improves performance in multi-source retrieval tasks by coordinating local and Web search agents and refining knowledge.

AI-generated summary

Recently, large reasoning models have demonstrated strong mathematical and coding abilities, and deep search leverages their reasoning capabilities in challenging information retrieval tasks. Existing deep search works are generally limited to a single knowledge source, either local or the Web. However, enterprises often require private deep search systems that can leverage search tools over both local and the Web corpus. Simply training an agent equipped with multiple search tools using flat reinforcement learning (RL) is a straightforward idea, but it has problems such as low training data efficiency and poor mastery of complex tools. To address the above issue, we propose a hierarchical agentic deep search framework, HierSearch, trained with hierarchical RL. At the low level, a local deep search agent and a Web deep search agent are trained to retrieve evidence from their corresponding domains. At the high level, a planner agent coordinates low-level agents and provides the final answer. Moreover, to prevent direct answer copying and error propagation, we design a knowledge refiner that filters out hallucinations and irrelevant evidence returned by low-level agents. Experiments show that HierSearch achieves better performance compared to flat RL, and outperforms various deep search and multi-source retrieval-augmented generation baselines in six benchmarks across general, finance, and medical domains.