---
title: "DynaKRAG：面向多跳检索增强生成的可学习证据控制统一框架"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2607.06507v1"
permalink: "https://feed.astralor.com/items/itm_a6c7e498cfad366d9418a77b"
published: "2026-07-07T17:09:36.000Z"
score: 77
channel: "ai"
tags: ["DynaKRAG","Qwen2.5-7B-Instruct","Retrieval-Augmented Generation","多跳问答","状态条件控制"]
---

# DynaKRAG：面向多跳检索增强生成的可学习证据控制统一框架

## AI 点评

> 这篇论文将多跳RAG的证据选择从固定流程提升到可学习的状态控制层面，思路非常优雅。对从事问答系统、检索增强生成的研究者尤其值得一读，特别是那些对模型自主决策能力感兴趣的人。

## 速览

多跳检索增强生成（RAG）需要顺序获取证据，现有方法通常采用固定流程或预定义控制拓扑，缺乏对证据操作的动态学习。DynaKRAG将证据获取建模为状态条件控制，通过可学习控制器选择原子操作（如检索、查询重写、证据评判、充分性判断），并动态更新证据状态。在HotpotQA等三个基准上，基于Qwen2.5-7B-Instruct的DynaKRAG取得最优F1分数，验证了协调检索、诊断与缺口导向获取的优势。

## 关键点

- 提出DynaKRAG框架，将多跳证据获取建模为状态条件控制，可学习选择证据操作。
- 定义原子操作集（检索、查询重写、证据评判、充分性判断），并通过有效性层构建可执行动作集。
- 在HotpotQA、2Wiki、MuSiQue上超越强基线，F1分别达0.5998、0.5340、0.3061。
- 消融实验表明学习控制器和充分性反馈对性能至关重要，额外检索并非总是有益。

## 全文

Multi-hop retrieval-augmented generation (RAG) acquires evidence sequentially, with each new document potentially revealing missing facts, bridge entities, query defects, or sufficient support for answering. Existing methods provide useful operations such as iterative retrieval, query reformulation, evidence critique, and sufficiency judging, but typically organize them within method-specific pipelines or predefined control topologies. This leaves underexplored how to learn a shared state-conditioned policy that chooses among currently valid evidence operations. We introduce DynaKRAG, which formulates multi-hop evidence acquisition as state-conditioned control over atomic evidence operations. At each step, a validity layer constructs the executable action set, and a learned controller selects the next operation. The resulting transition updates the evidence state and may enable new operations at subsequent steps. With Qwen2.5-7B-Instruct, DynaKRAG achieves F1 scores of 0.5998 on HotpotQA, 0.5340 on 2Wiki, and 0.3061 on MuSiQue, outperforming the strongest controlled baseline on all three benchmarks. Replacing the learned controller with a uniform-valid policy reduces F1 by 3.96--5.78 points, while removing sufficiency feedback hurts all three datasets. Controlled retrieval-cap experiments further show that additional retrieval is not uniformly beneficial. Together, these results demonstrate the benefit of coordinating retrieval, diagnosis, and gap-directed acquisition under an evolving evidence state.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_a6c7e498cfad366d9418a77b