AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

CVSBench: A Comprehensive Benchmark for Cross-view Spatial Reasoning and Dreaming

ruixun liu, Lingyu Zhang, Lanxuan Xue, Kaiyu Li, Bowen Fu, Xiangyong Cao

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Spatial reasoning in VLMs · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add inference-time CoT prompting (ablation)Qwen3-VL-4B · CVUSA-subset groundingmIoU (%): 3.68 -> 6.01Improved
add inference-time CoT prompting (ablation)Qwen3-VL-4B · FOV-subset G2Saccuracy (%): 26.70 -> 35.60Improved
add inference-time CoT prompting (ablation)Qwen3-VL-4B · FOV-subset View-Arrowaccuracy (%): 40.80 -> 21.90Hurt
add inference-time CoT prompting (ablation)Qwen3-VL-4B · FOV-subset View-Imageaccuracy (%): 27.10 -> 23.60Hurt
add inference-time CoT prompting (ablation)Qwen3-VL-4B · CVUSA-subset S2Gaccuracy (%): 62.40 -> 61.70Hurt
add Structured Scene CoT with SFT (ablation)Qwen3-VL-4B · FOV-subset G2S facadeaccuracy (%): 18.4 -> 41.3Improved
add Spatial Imagination CoT with SFT (ablation)Qwen3-VL-4B · CVUSA-subset G2S distanceaccuracy (%): 84.0 -> 89.1Improved
add Spatial Imagination CoT with SFT (ablation)Qwen3-VL-4B · CVUSA-subset G2S roofaccuracy (%): 63.2 -> 46.5Hurt
add RL after Structured Scene SFT (ablation)Qwen3-VL-4B · FOV-subset G2S facadeaccuracy (%): 41.3 -> 25.5Hurt
add RL after Structured Scene SFT (ablation)Qwen3-VL-4B · CVUSA-subset G2S connectionaccuracy (%): 28.1 -> 37.2Improved
add 3D miniature auxiliary view (ablation)Qwen3-VL-4B · FOV-subset S2G coloraccuracy (%): 24.5 -> 36.9Improved
add 3D miniature auxiliary view (ablation)Qwen3-VL-4B · FOV-subset G2S roofaccuracy (%): 9.9 -> 18.7Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy