InfoDeepSeek: Benchmarking Agentic Information Seeking for Retrieval-Augmented Generation

2025-05-21

AI Summary

Existing RAG benchmarks—relying on static corpora, fixed queries, and gold-document evaluation—fail to capture the dynamic information-seeking capabilities of agent-based RAG in open-web environments. Method: We introduce the first benchmark for intelligent web search evaluation tailored to dynamic Web settings, featuring a novel open-ended evaluation framework grounded in realistic webpage simulation. We propose a challenge question generation method balancing determinism, difficulty, and diversity; define a tri-dimensional fine-grained evaluation metric encompassing accuracy, utility, and conciseness; and integrate multi-LLM collaborative retrieval, scalable search trajectory modeling, and automated multi-dimensional assessment. Results: Extensive experiments across diverse LLMs, search engines, and query types reveal systematic agent behavioral differences, establish reproducible baselines, and yield transferable insights for optimizing agent-based RAG systems.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Data Mining & Knowledge Management: Knowledge Acquisition from the Web

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Retrieval-Augmented Generation (RAG) enhances large language models (LLMs) by grounding responses with retrieved information. As an emerging paradigm, Agentic RAG further enhances this process by introducing autonomous LLM agents into the information seeking process. However, existing benchmarks fall short in evaluating such systems, as they are confined to a static retrieval environment with a fixed, limited corpus} and simple queries that fail to elicit agentic behavior. Moreover, their evaluation protocols assess information seeking effectiveness by pre-defined gold sets of documents, making them unsuitable for the open-ended and dynamic nature of real-world web environments. To bridge this gap, we present InfoDeepSeek, a new benchmark with challenging questions designed for assessing agentic information seeking in real-world, dynamic web environments. We propose a systematic methodology for constructing challenging queries satisfying the criteria of determinacy, difficulty, and diversity. Based on this, we develop the first evaluation framework tailored to dynamic agentic information seeking, including fine-grained metrics about the accuracy, utility, and compactness of information seeking outcomes. Through extensive experiments across LLMs, search engines, and question types, InfoDeepSeek reveals nuanced agent behaviors and offers actionable insights for future research.

Problem

• Evaluating agentic RAG in dynamic web environments
• Lack of benchmarks for autonomous LLM agents
• Assessing information seeking accuracy and utility

Innovation

• Introduces autonomous LLM agents for dynamic RAG
• Develops challenging queries with determinacy and diversity
• Creates evaluation framework for dynamic agentic seeking