cs.CLMar 9, 2026

SPD-RAG: Sub-Agent Per Document Retrieval-Augmented Generation

Authors: Yagiz Can Akay, Muhammed Yusuf Kartal, Esra Alparslan, Faruk Ortakoyluoglu, Arda Akpinar

Organizations: TOBB University of Economics and Technology · OSTIM Technical University

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Answering complex, real-world queries often requires synthesizing facts scattered across vast document corpora. In these settings, standard retrieval-augmented generation (RAG) pipelines suffer from incomplete evidence coverage, while long-context large language models (LLMs) struggle to reason reliably over massive inputs. We introduce SPD-RAG, a hierarchical multi-agent framework for exhaustive cross-document question answering that decomposes the problem along the document axis. Each document is processed by a dedicated document-level agent operating only on its own content, enabling focused retrieval, while a coordinator dispatches tasks to relevant agents and aggregates their partial answers. Agent outputs are synthesized by merging partial answers through a token-bounded synthesis layer (which supports recursive map-reduce for massive corpora). This document-level specialization with centralized fusion improves scalability and answer quality in heterogeneous multidocument settings while yielding a modular, extensible retrieval pipeline. On the LOONG benchmark (EMNLP 2024) for long-context multi-document QA, SPD-RAG achieves an Avg Score of 58.1 (GPT-5 evaluation), outperforming Normal RAG (33.0) and Agentic RAG (32.8) while using only 38% of the API cost of a full-context baseline (68.0).

Figures & tables

Figure 1: Overview of SPD-RAG Architecture.
SystemAvg ScorePR (%)Avg Input TokAvg Total TokAvg Cost ($)Avg Latency (s)
Oracle Long-Context Baseline
Baseline (Full Context)68.031.4253,085255,3450.27345.6
RAG Baselines
Normal RAG33.013.722,17427,4300.08042.6
Agentic RAG32.88.881,87785,2900.09840.6
Our System
Table 1: Main results on our Loong evaluation set (102 instances: 40 academic paper, 62 financial report). Bold : best overall. Underline : best among RAG-based systems.
Figure 2: Comparison of Average Score across the four systems, broken down by task type (Spotlight Locating, Comparison, Clustering, and Chain of Reasoning).
SystemSpotlight Loc.ComparisonClusteringChain of Reasoning
 ( n=27n=27 )( n=15n=15 )( n=49n=49 )( n=11n=11 )
Baseline (Full Context)86.6 / 55.6%42.7 / 26.7%67.1 / 14.3%60.9 / 54.5%
Normal RAG69.7 / 37.0%37.7 / 26.7%15.4 / 0.0%14.8 / 0.0%
Agentic RAG73.4 / 22.2%23.3 / 20.0%16.7 / 0.0%17.9 / 0.0%
SPD-RAG74.2 / 44.4 %42.2 / 20.0%57.2 / 4.1%44.1 / 18.2%
Table 2: Avg Score / PR(%) by Loong task type. Each cell shows Avg Score / PR% .
SystemPaper ( n=40n=40 )Financial ( n=62n=62 )
Baseline78.8 / 30.0%61.0 / 32.3%
Normal RAG15.2 / 0.0%44.5 / 22.6%
Agentic RAG16.8 / 0.0%43.1 / 14.5%
SPD-RAG60.0 / 7.5%56.9 / 25.8%
Table 3: Avg Score / PR(%) by document domain.
Figure 3: Average score by document domain comparing the Baseline (Full Context), Normal RAG, Agentic RAG, and SPD-RAG systems.
SystemAvg ScoreAvg Cost ($)Score / $
Baseline68.00.273249.1
Normal RAG33.00.080412.5
Agentic RAG32.80.098334.7
SPD-RAG58.10.103564.1
Table 4: Cost–quality tradeoff.
Figure 4: Cost–Quality tradeoff. The scatter plot illustrates the Pareto frontier for multi-document QA.
Figure 5: Average per-query latency across systems.
More in this topic

Multi-Document QA

See all 20 papers

Explore similar work

CardsList
Aug 21, 2026

EnSI-RAG: Entity-Structure-Indexed Retrieval-Augmented Generation for Long-Document Question Answering

Question answering (QA) over long, connected documents remains challenging because relevant evidence may span multiple entities and their relationships. Existing retrieval-augmented generation (RAG) methods typically index documents as raw chunks and retrieve them through embedding similarity. Their performance degrades when chunk boundaries separate entities from supporting evidence or when a question requires multi-hop reasoning across the corpus. We propose EnSI-RAG (Entity-Structure-Indexed Retrieval-Augmented Generation), a framework that constructs a query-independent, entity-centered index. Each record (e, t, k, v) represents an entity e, its type t, a semantic category k in {property, relation, aspect}, and a value v, while retaining links to the original source passages. At query time, these records serve as retrieval handles, and an LLM synthesizes the retrieved passages into the final answer. This design separates evidence localization from answer synthesis while preserving traceable source evidence. Across Loong and Oolong, EnSI-RAG achieves an average accuracy of 78.24. Relative to the published baseline scores used as references, this is 6.62 points higher, suggesting its effectiveness across these settings. The code is available at https://github.com/RamonMeng/EnSI-RAG.

Xuanyu Meng, Jiashuo Sun, Jash Rajesh Parekh +1

University of Illinois Urbana-Champaign Urbana, Illinois, USA

Document QARetrieval-Augmented GenerationMulti-Hop QA
AskPDF
Apr 20, 2026

MASS-RAG: Multi-Agent Synthesis Retrieval-Augmented Generation

Large language models (LLMs) are widely used in retrieval-augmented generation (RAG) to incorporate external knowledge at inference time. However, when retrieved contexts are noisy, incomplete, or heterogeneous, a single generation process often struggles to reconcile evidence effectively. We propose MASS-RAG, a multi-agent synthesis approach to retrieval-augmented generation that structures evidence processing into multiple role-specialized agents. MASS-RAG applies distinct agents for evidence summarization, evidence extraction, and reasoning over retrieved documents, and combines their outputs through a dedicated synthesis stage to produce the final answer. This design exposes multiple intermediate evidence views, allowing the model to compare and integrate complementary information before answer generation. Experiments on four benchmarks show that MASS-RAG consistently improves performance over strong RAG baselines, particularly in settings where relevant evidence is distributed across retrieved contexts.

Xingchen Xiao, Heyan Huang, Runheng Liu +1

School of Computer Science and Technology, Beijing Institute of Technology · Department of Mathematical Sciences, Tsinghua University

Retrieval-Augmented GenerationMulti-Agent LLM SystemsMulti-Agent Reasoning
AskPDF
May 1, 2026

Hierarchical Abstract Tree for Cross-Document Retrieval-Augmented Generation

Retrieval-augmented generation (RAG) enhances large language models with external knowledge, and tree-based RAG organizes documents into hierarchical indexes to support queries at multiple granularities. However, existing Tree-RAG methods designed for single-document retrieval face critical challenges in scaling to cross-document multi-hop questions: (1) poor distribution adaptability, where kk-means clustering introduces noise due to rigid distribution assumptions; (2) structural isolation, as tree indexes lack explicit cross-document connections; and (3) coarse abstraction, which obscures fine-grained details. To address these limitations, we propose ΨΨ-RAG, a tree-RAG framework with two key components. First, a hierarchical abstract tree index built through an iterative "merging and collapse" process that adapts to data distributions without a priori assumption. Second, a multi-granular retrieval agent that intelligently interacts with the knowledge base with reorganized queries and an agent-powered hybrid retriever. ΨΨ-RAG supports diverse tasks from token-level question answering to document-level summarization. On cross-document multi-hop QA benchmarks, it outperforms RAPTOR by 25.9% and HippoRAG 2 by 7.4% in average F1 score. Code is available at https://github.com/Newiz430/Psi-RAG.

Ziwen Zhao, Menglin Yang

The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, China.

Multi-Hop QARetrieval-Augmented Generation
AskPDF