CoV: Chain-of-View Prompting for Spatial Reasoning

2026-01-08

arXiv.org

AI Summary

This work addresses the limitation of existing vision-language models in 3D embodied question answering, where fixed viewpoints hinder access to dispersed and occluded contextual information, thereby impeding complex spatial reasoning. The authors propose a training-free, test-time reasoning framework that dynamically selects and refines observation viewpoints through a coarse-to-fine active exploration mechanism to gather comprehensive spatial context. Their approach innovatively integrates chain-based viewpoint exploration with question-aligned anchor view selection, enabling a model-agnostic, open-ended viewpoint search strategy. A view selection agent filters redundant frames, while iterative reasoning combined with discrete camera actions efficiently acquires new observations. Evaluated on OpenEQA, the method improves LLM-Match by 11.56% on average (up to 13.62%) and demonstrates significant performance gains on ScanQA and SQA3D, showcasing strong test-time scalability.

Technology Category

• Computer Vision: 3D Computer Vision
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Intelligent Robots: Embodied AI

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Embodied question answering (EQA) in 3D environments often requires collecting context that is distributed across multiple viewpoints and partially occluded. However, most recent vision--language models (VLMs) are constrained to a fixed and finite set of input views, which limits their ability to acquire question-relevant context at inference time and hinders complex spatial reasoning. We propose Chain-of-View (CoV) prompting, a training-free, test-time reasoning framework that transforms a VLM into an active viewpoint reasoner through a coarse-to-fine exploration process. CoV first employs a View Selection agent to filter redundant frames and identify question-aligned anchor views. It then performs fine-grained view adjustment by interleaving iterative reasoning with discrete camera actions, obtaining new observations from the underlying 3D scene representation until sufficient context is gathered or a step budget is reached. We evaluate CoV on OpenEQA across four mainstream VLMs and obtain an average +11.56% improvement in LLM-Match, with a maximum gain of +13.62% on Qwen3-VL-Flash. CoV further exhibits test-time scaling: increasing the minimum action budget yields an additional +2.51% average improvement, peaking at +3.73% on Gemini-2.5-Flash. On ScanQA and SQA3D, CoV delivers strong performance (e.g., 116 CIDEr / 31.9 EM@1 on ScanQA and 51.1 EM@1 on SQA3D). Overall, these results suggest that question-aligned view selection coupled with open-view search is an effective, model-agnostic strategy for improving spatial reasoning in 3D EQA without additional training. Code is available on https://github.com/ziplab/CoV .

Problem

• Embodied Question Answering
• 3D environments
• spatial reasoning
• vision-language models
• viewpoint selection

Innovation

• Chain-of-View prompting
• spatial reasoning
• embodied question answering
• active viewpoint exploration
• vision-language models