WebWalker: Benchmarking LLMs in Web Traversal
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916888931467264 |
|---|---|
| author | Wu, Jialong Yin, Wenbiao Jiang, Yong Wang, Zhenglin Xi, Zekun Fang, Runnan Zhang, Linhai He, Yulan Zhou, Deyu Xie, Pengjun Huang, Fei |
| author_facet | Wu, Jialong Yin, Wenbiao Jiang, Yong Wang, Zhenglin Xi, Zekun Fang, Runnan Zhang, Linhai He, Yulan Zhou, Deyu Xie, Pengjun Huang, Fei |
| contents | Retrieval-augmented generation (RAG) demonstrates remarkable performance across tasks in open-domain question-answering. However, traditional search engines may retrieve shallow content, limiting the ability of LLMs to handle complex, multi-layered information. To address it, we introduce WebWalkerQA, a benchmark designed to assess the ability of LLMs to perform web traversal. It evaluates the capacity of LLMs to traverse a website's subpages to extract high-quality data systematically. We propose WebWalker, which is a multi-agent framework that mimics human-like web navigation through an explore-critic paradigm. Extensive experimental results show that WebWalkerQA is challenging and demonstrates the effectiveness of RAG combined with WebWalker, through the horizontal and vertical integration in real-world scenarios. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2501_07572 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | WebWalker: Benchmarking LLMs in Web Traversal Wu, Jialong Yin, Wenbiao Jiang, Yong Wang, Zhenglin Xi, Zekun Fang, Runnan Zhang, Linhai He, Yulan Zhou, Deyu Xie, Pengjun Huang, Fei Computation and Language Artificial Intelligence Retrieval-augmented generation (RAG) demonstrates remarkable performance across tasks in open-domain question-answering. However, traditional search engines may retrieve shallow content, limiting the ability of LLMs to handle complex, multi-layered information. To address it, we introduce WebWalkerQA, a benchmark designed to assess the ability of LLMs to perform web traversal. It evaluates the capacity of LLMs to traverse a website's subpages to extract high-quality data systematically. We propose WebWalker, which is a multi-agent framework that mimics human-like web navigation through an explore-critic paradigm. Extensive experimental results show that WebWalkerQA is challenging and demonstrates the effectiveness of RAG combined with WebWalker, through the horizontal and vertical integration in real-world scenarios. |
| title | WebWalker: Benchmarking LLMs in Web Traversal |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2501.07572 |