A2RAG: Adaptive Agentic Graph Retrieval for Cost-Aware and Reliable Reasoning
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866911405578387456 |
|---|---|
| author | Liu, Jiate Chen, Zebin Qiao, Shaobo Ju, Mingchen Zhang, Danting Han, Bocheng Yu, Shuyue Shu, Xin Wu, Jingling Wen, Dong Cao, Xin Liu, Guanfeng Yang, Zhengyi |
| author_facet | Liu, Jiate Chen, Zebin Qiao, Shaobo Ju, Mingchen Zhang, Danting Han, Bocheng Yu, Shuyue Shu, Xin Wu, Jingling Wen, Dong Cao, Xin Liu, Guanfeng Yang, Zhengyi |
| contents | Graph Retrieval-Augmented Generation (Graph-RAG) enhances multihop question answering by organizing corpora into knowledge graphs and routing evidence through relational structure. However, practical deployments face two persistent bottlenecks: (i) mixed-difficulty workloads where one-size-fits-all retrieval either wastes cost on easy queries or fails on hard multihop cases, and (ii) extraction loss, where graph abstraction omits fine-grained qualifiers that remain only in source text. We present A2RAG, an adaptive-and-agentic GraphRAG framework for cost-aware and reliable reasoning. A2RAG couples an adaptive controller that verifies evidence sufficiency and triggers targeted refinement only when necessary, with an agentic retriever that progressively escalates retrieval effort and maps graph signals back to provenance text to remain robust under extraction loss and incomplete graphs. Experiments on HotpotQA and 2WikiMultiHopQA demonstrate that A2RAG achieves +9.9/+11.8 absolute gains in Recall@2, while cutting token consumption and end-to-end latency by about 50% relative to iterative multihop baselines. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_21162 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | A2RAG: Adaptive Agentic Graph Retrieval for Cost-Aware and Reliable Reasoning Liu, Jiate Chen, Zebin Qiao, Shaobo Ju, Mingchen Zhang, Danting Han, Bocheng Yu, Shuyue Shu, Xin Wu, Jingling Wen, Dong Cao, Xin Liu, Guanfeng Yang, Zhengyi Information Retrieval Artificial Intelligence Databases Graph Retrieval-Augmented Generation (Graph-RAG) enhances multihop question answering by organizing corpora into knowledge graphs and routing evidence through relational structure. However, practical deployments face two persistent bottlenecks: (i) mixed-difficulty workloads where one-size-fits-all retrieval either wastes cost on easy queries or fails on hard multihop cases, and (ii) extraction loss, where graph abstraction omits fine-grained qualifiers that remain only in source text. We present A2RAG, an adaptive-and-agentic GraphRAG framework for cost-aware and reliable reasoning. A2RAG couples an adaptive controller that verifies evidence sufficiency and triggers targeted refinement only when necessary, with an agentic retriever that progressively escalates retrieval effort and maps graph signals back to provenance text to remain robust under extraction loss and incomplete graphs. Experiments on HotpotQA and 2WikiMultiHopQA demonstrate that A2RAG achieves +9.9/+11.8 absolute gains in Recall@2, while cutting token consumption and end-to-end latency by about 50% relative to iterative multihop baselines. |
| title | A2RAG: Adaptive Agentic Graph Retrieval for Cost-Aware and Reliable Reasoning |
| topic | Information Retrieval Artificial Intelligence Databases |
| url | https://arxiv.org/abs/2601.21162 |