Elamur Research⌕
☰
home / papers / 2026 / WebAggregator: Enhancing Compositional Reasoning Capabilities of Deep Research Agent Foundation Models

WebAggregator: Enhancing Compositional Reasoning Capabilities of Deep Research Agent Foundation Models

Paper · 2026 · Rui Wang, Ce Zhang, Jun-Yu Ma et al.

Summary

Deep Research agents need more than retrieval. They must combine scattered, heterogeneous evidence through multi-step reasoning. Existing training and evaluation resources mostly reward locating a single entity, so agents learn to call tools often instead of reasoning carefully.

The authors built WebAggregator, an automated, verifiable pipeline for generating data. A Proactive Explorer agent starts from anchor URLs and gathers linked material from real web environments, including search results, static and dynamic pages, files and images. A Compositional Logic Proposer then turns this material into questions that require aggregation. It follows a taxonomy of composition operations covering Element, Set, Scientific Analysis and Temporal Reasoning.

The resulting corpus, WebAggregatorQA, was filtered with rejection sampling and used for SFT of the WebAggregator models. The 32B model beats GPT-4.1 and is roughly on par with Claude-3.7-Sonnet on GAIA, WebWalkerQA and XBench. Fine-tuning also leads to deeper reasoning with fewer redundant tool calls.

A manually annotated test set shows that strong models fail even when all references are supplied. This points to compositional reasoning, not retrieval, as the main bottleneck.

Key points

  • Data scale: about 10K verifiable QA pairs drawn from roughly 50K websites. Seeds came from 5,000 queries, which yielded over 160,000 anchor links across 11+ domains.
  • Composition guidelines: questions follow 12 fine-grained subtypes grouped into four families (Element, Set, Scientific Analysis, Temporal).
  • Motivation from prior benchmarks: 30.79% of WebWalkerQA tasks and 43.2% of TaskCraft tasks need only simple entity localization.
  • Benchmark difficulty: on the WebAggregatorQA test set, Claude-3.7-Sonnet scores 28.3%. With full reference context provided, it reaches only 42.1%. Errors include faulty composition, accumulated rounding errors and failed self-correction.
  • Availability: code, data and models are announced as open source (github.com/Tencent/WebAggregator). The authors report zero overlap with existing benchmarks.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🏢TencentShenzhen · China134 works · 19,743 citations🎓Chinese University of Hong KongHong Kong · China126 works · 14,073 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksComputer Vision and Pattern Recognition · 2,869 worksInformation Systems · 1,437 works

🧭 Topics 1Research subjects and languages the work is about.Reasoning · 1,426 works

Cite

Cite as Rui Wang, Ce Zhang, Jun-Yu Ma, Jianshu Zhang, Hongru Wang, Yi Chen, Boyang Xue, Tianqing Fang, Zhisong Zhang, Hongming Zhang, Haitao Mi, Dong Yu and Kam-Fai Wong. 2026. WebAggregator: Enhancing Compositional Reasoning Capabilities of Deep Research Agent Foundation Models. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.1124

@inproceedings{wang2026webaggregator,
  title = {{WebAggregator: Enhancing Compositional Reasoning Capabilities of Deep Research Agent Foundation Models}},
  author = {Rui Wang, Ce Zhang, Jun-Yu Ma, Jianshu Zhang, Hongru Wang, Yi Chen, Boyang Xue, Tianqing Fang, Zhisong Zhang, Hongming Zhang, Haitao Mi, Dong Yu and Kam-Fai Wong},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.1124},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1124}
}

References

47 references, 18 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Carl Bereiter and Marlene Scardamalia. 1987. The psychology of written composition. find
  2. Yingshan Chang, Guihong Cao, Mridu Narang, Jianfeng Gao, Hisami Suzuki, and Yonatan Bisk. 2022. Webqa: Multihop and multimodal QA. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022, New Orleans, LA, USA, June 18-24, 2022, pages 16474–16483. IEEE. find
  3. Kaiyuan Chen, Yixin Ren, Yang Liu, Xiaobo Hu, Haotong Tian, Tianbao Xie, Fangfu Liu, Haoye Zhang, Hongzhang Liu, Yuan Gong, Chen Sun, Han Hou, Hui Yang, James Pan, Jianan Lou, Jiayi Mao, Jizheng Liu, Jinpeng Li, Kangyi Liu, and 14 others. 2025. xbench: Tracking agents productivity scaling with profe… — [arXiv:2506.13651](https://arxiv.org/abs/2506.13651) find
  4. Zehui Chen, Kuikun Liu, Qiuchen Wang, Wenwei Zhang, Jiangning Liu, Dahua Lin, Kai Chen, and Feng Zhao. 2024. Agent-FLAN: Designing data and methods of effective agent tuning for large language models. In Findings of the Association for Computational Linguistics: ACL 2024, pages 9354–9366, Bangkok, T…
  5. Xiang Deng, Yu Gu, Boyuan Zheng, Shijie Chen, Samuel Stevens, Boshi Wang, Huan Sun, and Yu Su. 2023. Mind2web: Towards a generalist agent for the web. Preprint, arXiv:2306.06070.
  6. Tianqing Fang, Zeming Chen, Yangqiu Song, and Antoine Bosselut. 2024. Complex reasoning over logical queries on commonsense knowledge graphs. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), ACL 2024, Bangkok, Thailand, August 11-16,… find
  7. Tianqing Fang, Hongming Zhang, Zhisong Zhang, Kaixin Ma, Wenhao Yu, Haitao Mi, and Dong Yu. 2025a. Webevolver: Enhancing web agent self-improvement with co-evolving world model. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 8970–8986.
  8. Tianqing Fang, Zhisong Zhang, Xiaoyang Wang, Rui Wang, Can Qin, Yuxuan Wan, Jun-Yu Ma, Ce Zhang, Jiaqi Chen, Xiyun Li, Hongming Zhang, Haitao Mi, and Dong Yu. 2025b. Cognitive kernel-pro: A framework for deep research agents and agent foundation models training. Preprint, arXiv:2508.00414. — [arXiv:2508.00414](https://arxiv.org/abs/2508.00414) find
  9. Yangyi Fang, Jiaye Lin, Xiaoliang Fu, Cong Qin, Haolin Shi, Chang Liu, and Peilin Zhao. 2026. Proximity-based multi-turn optimization: Practical credit assignment for llm agent training. arXiv preprint arXiv:2602.19225. find
  10. Lang Feng, Zhenghai Xue, Tingcong Liu, and Bo An. 2025. Group-in-group policy optimization for llm agent training. arXiv preprint arXiv:2505.10978. find
  11. Linda S. Flower and J. R. Hayes. 1981. A cognitive process theory of writing. College Composition & Communication. find
  12. Xiaoliang Fu, Jiaye Lin, Yangyi Fang, Binbin Zheng, Chaowen Hu, Zekai Shao, Cong Qin, Lu Pan, Ke Zeng, and Xunliang Cai. 2026. Maspo: Unifying gradient utilization, probability mass, and signal reliability for robust and sample-efficient llm reasoning. arXiv preprint arXiv:2602.17550.
  13. Hongliang He, Wenlin Yao, Kaixin Ma, Wenhao Yu, Yong Dai, Hongming Zhang, Zhenzhong Lan, and Dong Yu. 2024. WebVoyager: Building an end-to-end web agent with large multimodal models. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), p…
  14. Qiao Jin, Bhuwan Dhingra, Zhengping Liu, William W. Cohen, and Xinghua Lu. 2019. Pubmedqa: A dataset for biomedical research question answering. Preprint, arXiv:1909.06146.
  15. Satyapriya Krishna, Kalpesh Krishna, Anhad Mohananey, Steven Schwarcz, Adam Stambler, Shyam Upadhyay, and Manaal Faruqui. 2025. Fact, fetch, and reason: A unified evaluation of retrieval-augmented generation. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Associa… find
  16. Kuan Li, Zhongwang Zhang, Huifeng Yin, Liwen Zhang, Litu Ou, Jialong Wu, Wenbiao Yin, Baixuan Li, Zhengwei Tao, Xinyu Wang, Weizhou Shen, Junkai Zhang, Dingchu Zhang, Xixi Wu, Yong Jiang, Ming Yan, Pengjun Xie, Fei Huang, and Jingren Zhou. 2025a. WebSailor: Navigating Superhuman Reasoning for Web Ag… find
  17. Mukai Li, Qingcheng Zeng, Tianqing Fang, Zhenwen Liang, Linfeng Song, Qi Liu, Haitao Mi, and Dong Yu. 2026. Verified critical step optimization for llm agents. arXiv preprint arXiv:2602.03412. find
  18. Xiaoxi Li, Jiajie Jin, Guanting Dong, Hongjin Qian, Yutao Zhu, Yongkang Wu, Ji-Rong Wen, and Zhicheng Dou. 2025b. Webthinker: Empowering large reasoning models with deep research capability. Preprint, arXiv:2504.21776. find
  19. Zijian Li, Xin Guan, Bo Zhang, Shen Huang, Houquan Zhou, Shaopeng Lai, Ming Yan, Yong Jiang, Pengjun Xie, Fei Huang, Jun Zhang, and Jingren Zhou. 2025c. Webweaver: Structuring web-scale evidence with dynamic outlines for open-ended deep research. Preprint, arXiv:2509.13312. — [arXiv:2509.13312](https://arxiv.org/abs/2509.13312) find
  20. Grégoire Mialon, Clémentine Fourrier, Craig Swift, Thomas Wolf, Yann LeCun, and Thomas Scialom. 2023. Gaia: a benchmark for general ai assistants. Preprint, arXiv:2311.12983.
  21. Jifeng Dai, Xuan Dong, and 1 others. 2025. Mirothinker: Pushing the performance boundaries of open-source research agents via model, context, and interactive scaling. arXiv preprint arXiv:2511.11793. — [arXiv:2511.11793](https://arxiv.org/abs/2511.11793) find
  22. Monica.Im. 2025. Manus ai. Technical report, Monica.Im. find
  23. Tongyi. 2025. Tongyi-deepresearch. https://github.com/Alibaba-NLP/DeepResearch. ↗
  24. OpenAI. 2025. Introducing deep research \| OpenAI. find
  25. Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, and Ashish Sabharwal. 2022. MuSiQue: Multihop questions via single-hop question composition. Transactions of the Association for Computational Linguistics.
  26. Zile Qiao, Guoxin Chen, Xuanzhong Chen, Donglei Yu, Wenbiao Yin, Xinyu Wang, Zhen Zhang, Baixuan Li, Huifeng Yin, Kuan Li, Rui Min, Minpeng Liao, Yong Jiang, Pengjun Xie, Fei Huang, and Jingren Zhou. 2025. Webresearcher: Unleashing unbounded reasoning capability in long-horizon agents. Preprint, arX… — [arXiv:2509.13309](https://arxiv.org/abs/2509.13309) find
  27. Jiahao Qiu, Xuan Qi, Tongcheng Zhang, Xinzhe Juan, Jiacheng Guo, Yifu Lu, Yimin Wang, Zixin Yao, Qihan Ren, Xun Jiang, Xing Zhou, Dongrui Liu, Ling Yang, Yue Wu, Kaixuan Huang, Shilong Liu, Hongru Wang, and Mengdi Wang. 2025. Alita: Generalist agent enabling scalable agentic reasoning with minimal p… — [arXiv:2505.20286](https://arxiv.org/abs/2505.20286) find
  28. Aymeric Roucher, Albert Villanova del Moral, Thomas Wolf, Leandro von Werra, and Erik Kaunismäki. 2025. ‘smolagents‘: a smol library to build great agentic systems. https://github.com/hugging face/smolagents. ↗
  29. Priyanka Sen, Alham Fikri Aji, and Amir Saffari. 2022. Mintaka: A complex, natural, and multilingual dataset for end-to-end question answering. In Proceedings of the 29th International Conference on Computational Linguistics, pages 1604–1619, Gyeongju, Republic of Korea. International Committee on C… find
  30. Dingfeng Shi, Jingyi Cao, Qianben Chen, Weichen Sun, Weizhen Li, Hongxuan Lu, Fangchen Dong, Tianrui Qin, King Zhu, Minghao Yang, Jian Yang, Ge Zhang, Jiaheng Liu, Changwang Zhang, Jun Wang, Yuchen Eleanor Jiang, and Wangchunshu Zhou. 2025. TaskCraft: Automated Generation of Agentic Tasks. arXiv pre… — [arXiv:2506.10055](https://arxiv.org/abs/2506.10055) find
  31. Alon Talmor, Ori Yoran, Amnon Catav, Dan Lahav, Yizhong Wang, Akari Asai, Gabriel Ilharco, Hannaneh Hajishirzi, and Jonathan Berant. 2021. MUL-TIMODALQA: COMPLEX QUESTION ANSWER-ING OVER TEXT, TABLES AND IMAGES. find
  32. Zhengwei Tao, Jialong Wu, Wenbiao Yin, Junkai Zhang, Baixuan Li, Haiyang Shen, Kuan Li, Liwen Zhang, Xinyu Wang, Yong Jiang, Pengjun Xie, Fei Huang, and Jingren Zhou. 2025. Webshaper: Agentically data synthesizing via information-seeking formalization. Preprint, arXiv:2507.15061. — [arXiv:2507.15061](https://arxiv.org/abs/2507.15061) find
  33. MiroMind Team, Song Bai, Lidong Bing, Carson Chen, Guanzheng Chen, Yuntao Chen, Zhe Chen, Ziyi Chen, Yuxuan Wan, Tianqing Fang, Zaitang Li, Yintong Huo, Wenxuan Wang, Haitao Mi, Dong Yu, and Michael R Lyu. 2026. Inference-time scaling of verification: Self-evolving deep research agents via test-time…
  34. Tianyi Wang, Yixia Li, Long Li, Yibiao Chen, Shaohan Huang, Yun Chen, Peng Li, Yang Liu, and Guanhua Chen. 2026. Sppo: Sequence-level ppo for long-horizon reasoning tasks. Preprint, arXiv:2604.08865.
  35. Jason Wei, Zhiqing Sun, Spencer Papay, Scott McKinney, Jeffrey Han, Isa Fulford, Hyung Won Chung, Alex Tachard Passos, William Fedus, and Amelia Glaese. 2025. Browsecomp: A simple yet challenging benchmark for browsing agents. Preprint, arXiv:2504.12516. find
  36. Ryan Wong, Jiawei Wang, Junjie Zhao, Li Chen, Yan Gao, Long Zhang, Xuan Zhou, Zuo Wang, Kai Xiang, Ge Zhang, Wenhao Huang, Yang Wang, and Ke Wang. 2025. Widesearch: Benchmarking agentic broad info-seeking. Preprint, arXiv:2508.07999.
  37. Jialong Wu, Baixuan Li, Runnan Fang, Wenbiao Yin, Liwen Zhang, Zhengwei Tao, Dingchu Zhang, Zekun Xi, Yong Jiang, Pengjun Xie, Fei Huang, and Jingren Zhou. 2025a. WebDancer: Towards Autonomous Information Seeking Agency. arXiv preprint. ArXiv:2505.22648 [cs].
  38. Jialong Wu, Wenbiao Yin, Yong Jiang, Zhenglin Wang, Zekun Xi, Runnan Fang, Linhai Zhang, Yulan He, Deyu Zhou, Pengjun Xie, and Fei Huang. 2025b. Webwalker: Benchmarking llms in web traversal. Preprint, arXiv:2501.07572.
  39. Xianjie Wu, Jian Yang, Linzheng Chai, Ge Zhang, Jiaheng Liu, Xeron Du, Di Liang, Daixin Shu, Xianfu Cheng, Tianzhen Sun, and 1 others. 2025c. Tablebench: A comprehensive and complex benchmark for table question answering. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 39, p… find
  40. Ziyi Xia, Kun Luo, Hongjin Qian, and Zheng Liu. 2025. Open data synthesis for deep research. Preprint, arXiv:2509.00375. — [arXiv:2509.00375](https://arxiv.org/abs/2509.00375) find
  41. Yiheng Xu, Dunjie Lu, Zhennan Shen, Junli Wang, Zekun Wang, Yuchen Mao, Caiming Xiong, and Tao Yu. 2025. Agenttrek: Agent trajectory synthesis via guiding replay with web tutorials. Preprint, arXiv:2412.09605. — [arXiv:2412.09605](https://arxiv.org/abs/2412.09605) find
  42. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, and 41 others. 2025. Qwen3 technical report. arXiv preprint arXiv:2505.09388.
  43. 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 Yang, Jingren Zhou, Junyang Lin, Kai Dang, and 22 others. 2024. Qwen2.5 technical report. arXiv preprint arXi… find
  44. Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William Cohen, Ruslan Salakhutdinov, and Christopher D. Manning. 2018. HotpotQA: A dataset for diverse, explainable multi-hop question answering. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 2369…
  45. Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao. 2023. React: Synergizing reasoning and acting in language models. Preprint, arXiv:2210.03629.
  46. Jianshu Zhang, Dongyu Yao, Renjie Pi, Paul Pu Liang, and Yi R. Fung. 2025. Vlm2-bench: A closer look at how well vlms implicitly link explicit matching visual cues. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), ACL 2025, Vienna, A…
  47. Zhiwei Zhang, Fei Zhao, Rui Wang, Zezhong Wang, Bin Liang, Jiakang Wang, Yao Hu, Shaosheng Cao, and Kam-Fai Wong. 2026. Robust tool use via fission-grpo: Learning to recover from execution errors. arXiv preprint arXiv:2601.15625.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2026-webaggregator-enhancing-compositional-reasoning-capabilities/