We built a point-in-time financial deep research benchmark, featuring questions and rubrics generated through a rigorous, quality-controlled data pipeline. Additionally, we contracted financial experts to validate our data, with each spending an average of 1.2 hours on this meticulous review process. Leading LLMs such Opus-5 with our Finance Harness only score 44.9% on our leaderboard, showcasing the significant challenge our benchmark presents. We invite everyone to contribute to our leaderboard!</p>\n","updatedAt":"2026-08-06T21:52:14.385Z","author":{"_id":"64c04feeb746fe51543c1b7a","avatarUrl":"/avatars/77b87fca59a64c2463d652d454d78b13.svg","fullname":"Han","name":"Rujun","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8933539390563965},"editors":["Rujun"],"editorAvatarUrls":["/avatars/77b87fca59a64c2463d652d454d78b13.svg"],"reactions":[],"isReport":false}},{"id":"6a753a7244433412ccf79321","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":378,"isUserFollowing":false},"createdAt":"2026-08-07T01:52:50.000Z","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* [FinResearchBench II: A Deep Research Benchmark with Consensus-Derived Gold Rubrics for Distinguishing Financial Report Quality](https://huggingface.co/papers/2607.12252) (2026)\n* [FORCE-Bench: A Benchmark, Dataset, and Evaluation Harness for Agentic AI in Enterprise Finance](https://huggingface.co/papers/2607.19409) (2026)\n* [ICBCBench: An Industry Consortium Benchmark for Financial Deep Research](https://huggingface.co/papers/2606.17458) (2026)\n* [Deep Research in Physical Sciences: A Multi-Agent Framework and Comprehensive Benchmark](https://huggingface.co/papers/2606.18648) (2026)\n* [DRFLOW: A Deep Research Benchmark for Personalized Workflow Prediction](https://huggingface.co/papers/2606.18191) (2026)\n* [IPO Finance Agent: Benchmark of LLM Financial Analysts Beyond Finance Agent v2, with Automated Rubric Generation, on the SpaceX (SPCX) IPO](https://huggingface.co/papers/2606.23032) (2026)\n* [How Do Tool-Augmented LLM Agents Perform on Real-World Energy Analytics Tasks?](https://huggingface.co/papers/2606.26346) (2026)\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":"<p>This is an automated message from the <a href=\"https://huggingface.co/librarian-bots\">Librarian Bot</a>. I found the following papers similar to this paper. </p>\n<p>The following papers were recommended by the Semantic Scholar API </p>\n<ul>\n<li><a href=\"https://huggingface.co/papers/2607.12252\">FinResearchBench II: A Deep Research Benchmark with Consensus-Derived Gold Rubrics for Distinguishing Financial Report Quality</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2607.19409\">FORCE-Bench: A Benchmark, Dataset, and Evaluation Harness for Agentic AI in Enterprise Finance</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2606.17458\">ICBCBench: An Industry Consortium Benchmark for Financial Deep Research</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2606.18648\">Deep Research in Physical Sciences: A Multi-Agent Framework and Comprehensive Benchmark</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2606.18191\">DRFLOW: A Deep Research Benchmark for Personalized Workflow Prediction</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2606.23032\">IPO Finance Agent: Benchmark of LLM Financial Analysts Beyond Finance Agent v2, with Automated Rubric Generation, on the SpaceX (SPCX) IPO</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2606.26346\">How Do Tool-Augmented LLM Agents Perform on Real-World Energy Analytics Tasks?</a> (2026)</li>\n</ul>\n<p> Please give a thumbs up to this comment if you found it helpful!</p>\n<p> If you want recommendations for any Paper on Hugging Face checkout <a href=\"https://huggingface.co/spaces/librarian-bots/recommend_similar_papers\">this</a> Space</p>\n<p> You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: <code>@librarian-bot recommend</code></p>\n","updatedAt":"2026-08-07T01:52:50.411Z","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":378,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7345870137214661},"editors":["librarian-bot"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2607.27853","authors":[{"_id":"6a74ee80e1228e04b323809d","name":"Yijia Xiao","hidden":false},{"_id":"6a74ee80e1228e04b323809e","name":"Rujun Han","hidden":false},{"_id":"6a74ee80e1228e04b323809f","name":"Yanfei Chen","hidden":false},{"_id":"6a74ee80e1228e04b32380a0","name":"Zifeng Wang","hidden":false},{"_id":"6a74ee80e1228e04b32380a1","name":"Ke Jiang","hidden":false},{"_id":"6a74ee80e1228e04b32380a2","name":"Zhongying CuiZhu","hidden":false},{"_id":"6a74ee80e1228e04b32380a3","name":"Vishy Tirumalashetty","hidden":false},{"_id":"6a74ee80e1228e04b32380a4","name":"Wei Wang","hidden":false},{"_id":"6a74ee80e1228e04b32380a5","name":"Burak Gokturk","hidden":false},{"_id":"6a74ee80e1228e04b32380a6","name":"Tomas Pfister","hidden":false},{"_id":"6a74ee80e1228e04b32380a7","name":"Chen-Yu Lee","hidden":false}],"publishedAt":"2026-07-30T00:00:00.000Z","submittedOnDailyAt":"2026-08-06T00:00:00.000Z","title":"FinanceHarness: Autonomous Financial Deep Research Framework","submittedOnDailyBy":{"_id":"64c04feeb746fe51543c1b7a","avatarUrl":"/avatars/77b87fca59a64c2463d652d454d78b13.svg","isPro":false,"fullname":"Han","user":"Rujun","type":"user","name":"Rujun"},"summary":"Powered by advances in LLMs and autonomous agents, deep research has become one of the most widely adopted agentic products. However, most deep research systems write general-purpose reports, which are inadequate for financial deep research. Financial research demands specialized knowledge to analyze historical patterns and forecast upcoming events. Automating financial deep research therefore requires both a layered harness to drive the research agent and a verifiable, point-in-time benchmark that prevents leakage of future information. We present FinanceHarness, a harness that runs finance-oriented tools and practitioner-guided workflows, automating financial deep research end to end: environment and data construction, the agent execution loop, and reward modeling. We further propose FinanceGym, comprising thesis-driven research questions and rubrics that combine pre-cutoff and post-cutoff criteria. Professional expert validation yields an 82% pass rate. Even leading LLMs and agents score below 40% on the rubrics, showing that FinanceGym is challenging and leaves substantial headroom. With the same open-weight backbone, FinanceHarness improves the overall rubric score from 25.3% to 32.4%. FinanceHarness is available at https://github.com/Yijia-Xiao/FinanceHarness.","upvotes":3,"discussionId":"6a74ee81e1228e04b32380a8","projectPage":"https://financegym.github.io/"},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"64c04feeb746fe51543c1b7a","avatarUrl":"/avatars/77b87fca59a64c2463d652d454d78b13.svg","isPro":false,"fullname":"Han","user":"Rujun","type":"user"},{"_id":"64d1fe784dfd5df7076bfd1a","avatarUrl":"/avatars/e2bb108d1f6b1383f0e3f263c4747b3d.svg","isPro":false,"fullname":"Chen-Yu Lee","user":"chenyulee","type":"user"},{"_id":"67c772e077464ecefff08038","avatarUrl":"/avatars/23ad9716715bbaf934a646e45c8f5ae0.svg","isPro":false,"fullname":"Zoey CuiZhu","user":"cuizz11","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"query":{}}">

Cite arxiv.org/abs/2607.27853 in a model README.md to link it from this page.

Cite arxiv.org/abs/2607.27853 in a Space README.md to link it from this page.

Add this paper to a collection to link it from this page.