面向金融、法律与医疗文档的自适应查询路由：分层混合检索框架

View PDF HTML (experimental)

Abstract:Retrieval-Augmented Generation (RAG) has become the standard paradigm for grounding Large Language Model outputs in external knowledge. Lumer et al. presented the first systematic evaluation comparing vector-based agentic RAG against hierarchical node-based reasoning systems for financial document QA across 1,200 SEC filings, finding vector-based systems achieved a 68% win rate. Concurrently, the PageIndex framework reported 98.7% accuracy on FinanceBench through purely reasoning-based retrieval. This paper extends their work by: (i) implementing and evaluating three retrieval configurations: Vector RAG, Tree Reasoning, and the proposed Adaptive Hybrid Retrieval (AHR), on a small controlled corpus of financial, legal, and medical documents; (ii) introducing a four-tier query complexity benchmark; and (iii) employing LLM-as-judge evaluation with GPT-4o-mini. Experiments reveal that Tree Reasoning achieves the highest overall score (0.900), but no single paradigm dominates across all tiers: AHR scores highest on cross-reference queries (0.850), all three methods tie on multi-section queries (0.929), and Vector RAG scores highest on the single multi-document query (0.900). Vector RAG has the lowest mean latency (1.61 s against 3.40 s for Tree Reasoning). On the first 50 publicly released FinanceBench questions, using the evidence snippets supplied with the benchmark as the retrieval corpus, Tree Reasoning scores 0.938, Hybrid AHR 0.901, and Vector RAG 0.821. Because the configurations also differ in document access, generation prompts and output length limits, these differences cannot be attributed to retrieval architecture alone. In these preliminary results the adaptive router does not outperform fixed tree reasoning overall (0.845 vs. 0.900); they motivate a larger evaluation of when query-based routing helps, rather than establishing its benefit.

Comments:	12 pages, 4 figures, 6 tables. v2: method description corrected to match the experiment code; section-level retrieval metrics withdrawn because the original scorer ignored document identity; FinanceBench corpus clarified (supplied evidence snippets, first 50 public questions); prompt and output-length differences between configurations disclosed
Subjects:	Information Retrieval (cs.IR); Artificial Intelligence (cs.AI)
Cite as:	arXiv:2604.14222 [cs.IR]
	(or arXiv:2604.14222v2 [cs.IR] for this version)
	https://doi.org/10.48550/arXiv.2604.14222 arXiv-issued DOI via DataCite

Submission history

From: Afshan Hashmi Dr. [view email]
[v1] Tue, 14 Apr 2026 10:48:13 UTC (1,086 KB)
[v2] Wed, 28 Oct 2026 09:45:47 UTC (332 KB)