Reason Before You Retrieve: Agentic Planning for Multi-modal RAG

• URL: http://arxiv.org/abs/2607.22643v1
• Date: Wed, 24 Jun 2026 02:40:49 GMT
• Title: Reason Before You Retrieve: Agentic Planning for Multi-modal RAG
• Abstract summary: We propose MM-R2, a multimodal agentic retrieval framework that reasons before retrieval by explicitly modeling both what to retrieve and where to search.<n> MM-R2 first constructs an intent-grounded retrieval state from the image-question pair, capturing the information need, grounded referent, and retrieval constraints.<n>It then performs retrieval over a structured KnowledgeMap, where the agent selects relevant retrieval units before issuing grounded queries within them.<n>Experiments on Infoseek and Encyclopedic VQA datasets show that MM-R2 substantially outperforms strong baselines on answer accuracy while also yielding more interpretable and verifiable retrieval trajectories.
• Score: 37.890744636035144
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Multimodal retrieval-augmented generation (mRAG) aims to answer image-text queries with external knowledge, but most existing systems still retrieve directly from raw multimodal input over a flat evidence space. This design often struggles with two key challenges: the retrieval target is under-specified because the question intent must be grounded to the correct visual referent, and the search space is weakly structured, forcing semantically distinct evidence to compete in a single global ranking step. We propose MM-R2, a multimodal agentic retrieval framework that reasons before retrieval by explicitly modeling both what to retrieve and where to search. MM-R2 first constructs an intent-grounded retrieval state from the image-question pair, capturing the information need, grounded referent, and retrieval constraints. It then performs retrieval over a structured KnowledgeMap, where the agent selects relevant retrieval units before issuing grounded queries within them. To enable this capability, we build MM-R2-Traj, a large-scale trajectory dataset of multi-step retrieval processes, and adopt a two-stage post-training strategy with supervised fine-tuning and GRPO. Experiments on Infoseek and Encyclopedic VQA datasets show that MM-R2 substantially outperforms strong baselines on answer accuracy while also yielding more interpretable and verifiable retrieval trajectories.

Related papers

• Pixel-Grounded Retrieval for Knowledgeable Large Multimodal Models [58.46663983451155]
  PixSearch is an end-to-end Segmenting Large Multimodal Model (LMM) that unifies region-level perception and retrieval-augmented reasoning.<n>During encoding, PixSearch emits search> tokens to trigger retrieval, selects query modalities (text, image, or region), and generates pixel-level masks that directly serve as visual queries.<n>On egocentric and entity-centric VQA benchmarks, PixSearch substantially improves factual consistency and generalization.
  arXiv  Detail & Related papers  (2026-01-27T00:46:08Z)
• Recurrence Meets Transformers for Universal Multimodal Retrieval [59.92546492752452]
  ReT-2 is a unified retrieval model that supports multimodal queries composed of both images and text.<n>We evaluate ReT-2 on the challenging M2KR and M-BEIR benchmarks across different retrieval configurations.<n>When integrated into retrieval-augmented generation pipelines, ReT-2 also improves downstream performance on Encyclopedic-VQA and InfoSeek datasets.
  arXiv  Detail & Related papers  (2025-09-10T18:00:29Z)
• Fishing for Answers: Exploring One-shot vs. Iterative Retrieval Strategies for Retrieval Augmented Generation [11.180502261031789]
  Retrieval-Augmented Generation (RAG) based on Large Language Models (LLMs) is a powerful solution to understand and query the industry's closed-source documents.<n>However, basic RAG often struggles with complex QA tasks in legal and regulatory domains.<n>We explore two strategies to improve evidence coverage and answer quality.
  arXiv  Detail & Related papers  (2025-09-05T05:44:50Z)
• MMSearch-R1: Incentivizing LMMs to Search [49.889749277236376]
  We present MMSearch-R1, the first end-to-end reinforcement learning framework that enables on-demand, multi-turn search in real-world Internet environments.<n>Our framework integrates both image and text search tools, allowing the model to reason about when and how to invoke them guided by an outcome-based reward with a search penalty.
  arXiv  Detail & Related papers  (2025-06-25T17:59:42Z)
• Benchmarking Multimodal Retrieval Augmented Generation with Dynamic VQA Dataset and Self-adaptive Planning Agent [92.5712549836791]
  Multimodal Retrieval Augmented Generation (mRAG) plays an important role in mitigating the "hallucination" issue inherent in multimodal large language models (MLLMs)<n>We propose the first self-adaptive planning agent for multimodal retrieval, OmniSearch.
  arXiv  Detail & Related papers  (2024-11-05T09:27:21Z)
• End-to-end Knowledge Retrieval with Multi-modal Queries [50.01264794081951]
  ReMuQ requires a system to retrieve knowledge from a large corpus by integrating contents from both text and image queries. We introduce a retriever model ReViz'' that can directly process input text and images to retrieve relevant knowledge in an end-to-end fashion. We demonstrate superior performance in retrieval on two datasets under zero-shot settings.
  arXiv  Detail & Related papers  (2023-06-01T08:04:12Z)
• UniKGQA: Unified Retrieval and Reasoning for Solving Multi-hop Question Answering Over Knowledge Graph [89.98762327725112]
  Multi-hop Question Answering over Knowledge Graph(KGQA) aims to find the answer entities that are multiple hops away from the topic entities mentioned in a natural language question. We propose UniKGQA, a novel approach for multi-hop KGQA task, by unifying retrieval and reasoning in both model architecture and parameter learning.
  arXiv  Detail & Related papers  (2022-12-02T04:08:09Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.