Information seeking demands iterative evidence gathering and reflective reasoning, yet large language models (LLMs) still struggle with it in open-web question answering. Existing prompting and supervised fine-tuning (SFT) methods remain fixed by prompt rules or training corpora, and are usually benchmarked only on well-structured wiki sources, limiting real-world adaptability. We introduce WebPuzzle, a 24k-sample training and 275-sample test benchmark that evaluates information seeking on the live internet, across both wiki and open-domain queries. Leveraging 7k WebPuzzle instances, we develop DeepDiver, a reinforcement-learning (RL) framework that cultivates Search Intensity Scaling (SIS)-an emergent ability to escalate search frequency and depth instead of settling on overconfident, under-evidenced answers. With SIS, Qwen2.5-7B-Instruct and Pangu-7B-Reasoner attain performance on real-web tasks comparable to the 671B-parameter DeepSeek-R1. We detail DeepDiver's curriculum from cold-start SFT to a well designed RL procedure, and show that its seeking policy generalized from closed-ended queries to open-ended generation such as long-form writing. Our results advance adaptive information seeking in LLMs and provide a rigorous benchmark for future work.
@article{arxiv.2505.24332,
title = {DeepDiver: Adaptive Search Intensity Scaling via Open-Web Reinforcement Learning},
author = {Wenxuan Shi and Haochen Tan and Chuqiao Kuang and Xiaoguang Li and Xiaozhe Ren and Chen Zhang and Hanting Chen and Yasheng Wang and Lu Hou and Lifeng Shang},
journal= {arXiv preprint arXiv:2505.24332},
year = {2025}
}