Faithful Grounded Visual Reasoning via Learned Proxy-Tokens

• URL: http://arxiv.org/abs/2606.23354v1
• Date: Mon, 22 Jun 2026 13:54:48 GMT
• Title: Faithful Grounded Visual Reasoning via Learned Proxy-Tokens
• Abstract summary: Grounded Visual Reasoning (GVR) approaches attempt to improve interpretability by explicitly couple textual rationales with visual grounding information.<n>This mechanism lacks a learnable semantic link to the visual features, often resulting in a semantic-spatial gap.<n>We introduce Composer, a MLLM that leverages a novel visual grounding mechanism based on learned proxy-tokens.
• Score: 1.5749416770494706
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Multimodal Large Language Models (MLLMs) have achieved remarkable success in Visual Question Answering (VQA), yet their "black-box" nature hinders deployment in critical domains. Grounded Visual Reasoning (GVR) approaches attempt to improve interpretability by explicitly couple textual rationales with visual grounding information, which are typically textual coordinates. This mechanism lacks a learnable semantic link to the visual features, often resulting in a semantic-spatial gap where the model hallucinates coordinates that do not correspond to image evidences. In this work, we introduce Composer, a MLLM that leverages a novel visual grounding mechanism based on learned proxy-tokens to promote faithful interpretability. These discrete symbolic pointers explicitly index the image latent space, allowing the model to manipulate visual regions as addressable, semantically manipulable sets. To rigorously validate our novel grounding mechanism, we constructed ComposerGCoT, a dataset synthesized to enable holistic assessment of reasoning consistency and grounding accuracy. Experimental results indicate that Composer achieves performance parity with its coordinate-based counterpart in final answer accuracy, while improving visual grounding accuracy by +9.0 points. By demonstrating that discrete proxy-tokens capture spatial semantics more effectively than typical textual coordinates, we establish that visual grounding mechanisms with learnable semantic links represent a promising path toward trustworthy and reliable MLLMs.

Related papers

• USS: Unified Spatial-Semantic Prompts for Embodied Visual Tracking with Latent Dynamics Learning [18.003030497082623]
  Embodied Visual Tracking (EVT) requires an agent to continuously follow a specified target while actively moving through dynamic environments.<n>We propose a paradigm shift, reframing target indication in EVT from text-only specification to unified spatial-semantic prompting.<n>We introduce Unified Spatial-Semantic Prompts for Embodied Visual Tracking with Latent Dynamics Learning.
  arXiv  Detail & Related papers  (2026-06-24T14:25:05Z)
• AffordGrasp: Cross-Modal Diffusion for Affordance-Aware Grasp Synthesis [50.793806818677716]
  AffordGrasp produces physically stable and semantically faithful human grasps with high precision.<n>AffordGrasp integrates an affordance-aware latent representation of hand poses with a dual-conditioning diffusion process.<n>We evaluate AffordGrasp across four instruction-augmented benchmarks derived from HO-3D, OakInk, GRAB, and AffordPose.
  arXiv  Detail & Related papers  (2026-03-09T06:56:35Z)
• The Perceptual Observatory Characterizing Robustness and Grounding in MLLMs [44.71703930770065]
  We present The Perceptual Observatory, a framework that characterizes MLLMs across verticals like face matching and text-in-vision comprehension capabilities.<n>The Perceptual Observatory moves beyond leaderboard accuracy to yield insights into how MLLMs preserve perceptual grounding and relational structure under perturbations.
  arXiv  Detail & Related papers  (2025-12-17T20:22:23Z)
• SATGround: A Spatially-Aware Approach for Visual Grounding in Remote Sensing [57.609801041296095]
  Vision-language models (VLMs) are emerging as powerful tools for remote sensing.<n>We enhance VLM-based visual grounding in satellite imagery by proposing a novel structured localization mechanism.
  arXiv  Detail & Related papers  (2025-12-09T18:15:43Z)
• Mind-the-Glitch: Visual Correspondence for Detecting Inconsistencies in Subject-Driven Generation [120.23172120151821]
  We propose a novel approach for disentangling visual and semantic features from the backbones of pre-trained diffusion models.<n>We introduce an automated pipeline that constructs image pairs with annotated semantic and visual correspondences.<n>We propose a new metric, Visual Semantic Matching, that quantifies visual inconsistencies in subject-driven image generation.
  arXiv  Detail & Related papers  (2025-09-26T07:11:55Z)
• Generalized Decoupled Learning for Enhancing Open-Vocabulary Dense Perception [71.26728044621458]
  DeCLIP is a novel framework that enhances CLIP by decoupling the self-attention module to obtain content'' and context'' features respectively.<n>It consistently achieves state-of-the-art performance across a broad spectrum of tasks, including 2D detection and segmentation, 3D instance segmentation, video instance segmentation, and 6D object pose estimation.
  arXiv  Detail & Related papers  (2025-08-15T06:43:51Z)
• Mind the Gap: Benchmarking Spatial Reasoning in Vision-Language Models [14.442394137843923]
  We present a detailed analysis that first delineates the core elements of spatial reasoning.<n>We then assesses the performance of these models in both synthetic and real-world images.
  arXiv  Detail & Related papers  (2025-03-25T14:34:06Z)
• Towards Grounded Visual Spatial Reasoning in Multi-Modal Vision Language Models [3.86170450233149]
  We show that large vision-and-language models (VLMs) trained to match images with text lack fine-grained understanding of spatial relations. We propose an alternative fine-grained, compositional approach for recognizing and ranking spatial clauses.
  arXiv  Detail & Related papers  (2023-08-18T18:58:54Z)
• Sim-To-Real Transfer of Visual Grounding for Human-Aided Ambiguity Resolution [0.0]
  We consider the task of visual grounding, where the agent segments an object from a crowded scene given a natural language description. Modern holistic approaches to visual grounding usually ignore language structure and struggle to cover generic domains. We introduce a fully decoupled modular framework for compositional visual grounding of entities, attributes, and spatial relations.
  arXiv  Detail & Related papers  (2022-05-24T14:12:32Z)
• Cross-modal Representation Learning for Zero-shot Action Recognition [67.57406812235767]
  We present a cross-modal Transformer-based framework, which jointly encodes video data and text labels for zero-shot action recognition (ZSAR) Our model employs a conceptually new pipeline by which visual representations are learned in conjunction with visual-semantic associations in an end-to-end manner. Experiment results show our model considerably improves upon the state of the arts in ZSAR, reaching encouraging top-1 accuracy on UCF101, HMDB51, and ActivityNet benchmark datasets.
  arXiv  Detail & Related papers  (2022-05-03T17:39:27Z)
• Co-Grounding Networks with Semantic Attention for Referring Expression Comprehension in Videos [96.85840365678649]
  We tackle the problem of referring expression comprehension in videos with an elegant one-stage framework. We enhance the single-frame grounding accuracy by semantic attention learning and improve the cross-frame grounding consistency. Our model is also applicable to referring expression comprehension in images, illustrated by the improved performance on the RefCOCO dataset.
  arXiv  Detail & Related papers  (2021-03-23T06:42:49Z)

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.