Scene Graph Thinking: Reinforcing Structured Visual Reasoning for Multimodal Large Language Models

• URL: http://arxiv.org/abs/2607.05716v3
• Date: Mon, 13 Jul 2026 06:34:36 GMT
• Title: Scene Graph Thinking: Reinforcing Structured Visual Reasoning for Multimodal Large Language Models
• Abstract summary: Scene Graph Thinking (SaGe) is a novel paradigm that enables fine-grained and structured visual reasoning through explicit scene-graph representations.<n>We construct 120K high-quality training data by sampling reasoning traces from scene graphs.<n>With curated data and graph-aligned training, our approach achieves significant improvements across eight multimodal benchmarks.
• Score: 46.09809083890632
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Multimodal Large Language Models (MLLMs) have demonstrated strong perception and reasoning capabilities. However, most existing models focus on isolated objects and neglect structured relationships for efficient target navigation, limiting their performance on visually intensive tasks. To address this challenge, we introduce Scene Graph Thinking (SaGe), a novel paradigm that enables fine-grained and structured visual reasoning through explicit scene-graph representations. Specifically, we first introduce an automated data engine that converts flat image-text corpora into structured scene graphs, where hierarchical entities constitute the nodes and diverse visual relations define the edges. Building upon this, we construct 120K high-quality training data by sampling reasoning traces from scene graphs. Then, two-stage graph-aligned post-training paradigms are introduced, where supervised fine-tuning internalizes MLLMs with structured reasoning, and subsequent reinforcement fine-tuning proposes node-as-proxy graph rewards to consolidate efficient graph exploration. With curated data and graph-aligned training, our approach achieves significant improvements across eight multimodal benchmarks, demonstrating strong effectiveness on fine-grained perception and reasoning tasks. Code is available at https://github.com/zwyang6/SaGe.

Related papers

• AdaSTORM: Scaling LLM Reasoning on Dynamic Graphs via Adaptive Spatio-Temporal Multi-Agent Collaboration [17.873099309973657]
  This paper presents Scaling Reasoning on Dynamic Graphs via Adaptive Spatio-Temporal MultiAgent-Temporal Collaboration (AdaSTORM)<n>AdaSTORM is the first multi-agent framework tailored for dynamic graph reasoning.<n>It achieves state-of-the-art accuracy on existing benchmarks and generalizes robustly to real-world datasets.
  arXiv  Detail & Related papers  (2026-06-15T07:32:44Z)
• Edge-Aware Curvature Modeling for Graph Understanding in Large Language Models [22.713901071316915]
  Graph-aware Large Language Models (LLMs) have shown promising capabilities in jointly modeling graph-structured data and textual information.<n>We propose a CureLLM framework of Curvature-enhanced Graph Representations for Large Language Model.
  arXiv  Detail & Related papers  (2026-06-04T12:12:34Z)
• Semantic Generative Tuning for Unified Multimodal Models [62.18894352635965]
  Unified multimodal models (UMMs) strive to consolidate visual understanding and visual generation within a single architecture.<n> prevailing training paradigms independently optimize understanding via sparse text signals and generation through dense pixel objectives.<n>This work presents the first systematic investigation into generative post-training, where we formulate hierarchical visual tasks as generative proxies to bridge the isolation in UMMs.
  arXiv  Detail & Related papers  (2026-05-18T17:46:46Z)
• RAG-GFM: Overcoming In-Memory Bottlenecks in Graph Foundation Models via Retrieval-Augmented Generation [27.59455285600957]
  Graph Foundation Models (GFMs) have emerged as a frontier in graph learning, which are expected to deliver transferable representations across diverse tasks.<n>We propose RAG-GFM, a Retrieval-Augmented Generation aided Graph Foundation Model that offloads knowledge from parameters.<n>We show that RAG-GFM consistently outperforms 13 state-of-the-art baselines in both cross-domain node and graph classification.
  arXiv  Detail & Related papers  (2026-01-21T16:02:43Z)
• Revisiting Graph Neural Networks on Graph-level Tasks: Comprehensive Experiments, Analysis, and Improvements [54.006506479865344]
  We propose a unified evaluation framework for graph-level Graph Neural Networks (GNNs)<n>This framework provides a standardized setting to evaluate GNNs across diverse datasets.<n>We also propose a novel GNN model with enhanced expressivity and generalization capabilities.
  arXiv  Detail & Related papers  (2025-01-01T08:48:53Z)
• Can Graph Neural Networks Learn Language with Extremely Weak Text Supervision? [62.12375949429938]
  We propose a multi-modal prompt learning paradigm to adapt pre-trained Graph Neural Networks to downstream tasks and data.<n>Our new paradigm embeds the graphs directly in the same space as the Large Language Models (LLMs) by learning both graph prompts and text prompts simultaneously.<n>We build the first CLIP-style zero-shot classification prototype that can generalize GNNs to unseen classes with extremely weak text supervision.
  arXiv  Detail & Related papers  (2024-12-11T08:03:35Z)
• Towards Graph Foundation Models: Training on Knowledge Graphs Enables Transferability to General Graphs [26.477872205199667]
  We introduce SCR, a unified graph reasoning framework designed to train on knowledge graphs.<n>We propose semantic-conditioned message passing, a novel mechanism addressing the inherent semantic isolation in traditional KG reasoning.<n>Our results show substantial performance gains over existing foundation models.
  arXiv  Detail & Related papers  (2024-10-16T14:26:08Z)
• InstructG2I: Synthesizing Images from Multimodal Attributed Graphs [50.852150521561676]
  We propose a graph context-conditioned diffusion model called InstructG2I. InstructG2I first exploits the graph structure and multimodal information to conduct informative neighbor sampling. A Graph-QFormer encoder adaptively encodes the graph nodes into an auxiliary set of graph prompts to guide the denoising process.
  arXiv  Detail & Related papers  (2024-10-09T17:56:15Z)
• Exploring Graph Structure Comprehension Ability of Multimodal Large Language Models: Case Studies [7.067145619709089]
  This study investigates the impact of graph visualisations on Large Language Models (LLMs) performance. Our experiments compare the effectiveness of multimodal approaches against purely textual graph representations.
  arXiv  Detail & Related papers  (2024-09-13T14:26:58Z)
• GraphGPT: Graph Instruction Tuning for Large Language Models [27.036935149004726]
  Graph Neural Networks (GNNs) have evolved to understand graph structures. To enhance robustness, self-supervised learning (SSL) has become a vital tool for data augmentation. Our research tackles this by advancing graph model generalization in zero-shot learning environments.
  arXiv  Detail & Related papers  (2023-10-19T06:17:46Z)
• Structure-CLIP: Towards Scene Graph Knowledge to Enhance Multi-modal Structured Representations [70.41385310930846]
  We present an end-to-end framework Structure-CLIP to enhance multi-modal structured representations. We use scene graphs to guide the construction of semantic negative examples, which results in an increased emphasis on learning structured representations. A Knowledge-Enhance (KEE) is proposed to leverage SGK as input to further enhance structured representations.
  arXiv  Detail & Related papers  (2023-05-06T03:57:05Z)
• Scene Graph Modification as Incremental Structure Expanding [61.84291817776118]
  We focus on scene graph modification (SGM), where the system is required to learn how to update an existing scene graph based on a natural language query. We frame SGM as a graph expansion task by introducing the incremental structure expanding (ISE) We construct a challenging dataset that contains more complicated queries and larger scene graphs than existing datasets.
  arXiv  Detail & Related papers  (2022-09-15T16:26:14Z)

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.