RAGCap-Bench: Benchmarking Capabilities of LLMs in Agentic Retrieval Augmented Generation Systems

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Lin, Jingru, Zhang, Chen, Liu, Stephen Y., Li, Haizhou
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910243905077248
author Lin, Jingru
Zhang, Chen
Liu, Stephen Y.
Li, Haizhou
author_facet Lin, Jingru
Zhang, Chen
Liu, Stephen Y.
Li, Haizhou
contents Retrieval-Augmented Generation (RAG) mitigates key limitations of Large Language Models (LLMs)-such as factual errors, outdated knowledge, and hallucinations-by dynamically retrieving external information. Recent work extends this paradigm through agentic RAG systems, where LLMs act as agents to iteratively plan, retrieve, and reason over complex queries. However, these systems still struggle with challenging multi-hop questions, and their intermediate reasoning capabilities remain underexplored. To address this, we propose RAGCap-Bench, a capability-oriented benchmark for fine-grained evaluation of intermediate tasks in agentic RAG workflows. We analyze outputs from state-of-the-art systems to identify common tasks and the core capabilities required for their execution, then construct a taxonomy of typical LLM errors to design targeted evaluation questions. Experiments show that "slow-thinking" models with stronger RAGCap performance achieve better end-to-end results, underscoring the benchmark's validity and the importance of enhancing these intermediate capabilities.
format Preprint
id arxiv_https___arxiv_org_abs_2510_13910
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle RAGCap-Bench: Benchmarking Capabilities of LLMs in Agentic Retrieval Augmented Generation Systems
Lin, Jingru
Zhang, Chen
Liu, Stephen Y.
Li, Haizhou
Computation and Language
Retrieval-Augmented Generation (RAG) mitigates key limitations of Large Language Models (LLMs)-such as factual errors, outdated knowledge, and hallucinations-by dynamically retrieving external information. Recent work extends this paradigm through agentic RAG systems, where LLMs act as agents to iteratively plan, retrieve, and reason over complex queries. However, these systems still struggle with challenging multi-hop questions, and their intermediate reasoning capabilities remain underexplored. To address this, we propose RAGCap-Bench, a capability-oriented benchmark for fine-grained evaluation of intermediate tasks in agentic RAG workflows. We analyze outputs from state-of-the-art systems to identify common tasks and the core capabilities required for their execution, then construct a taxonomy of typical LLM errors to design targeted evaluation questions. Experiments show that "slow-thinking" models with stronger RAGCap performance achieve better end-to-end results, underscoring the benchmark's validity and the importance of enhancing these intermediate capabilities.
title RAGCap-Bench: Benchmarking Capabilities of LLMs in Agentic Retrieval Augmented Generation Systems
topic Computation and Language
url https://arxiv.org/abs/2510.13910