AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Aligning Cross-View Visual Geometries in LVLMs Through Human-Like Reasoning Learning

Yuming Qiao, Liang Luo, Dan Meng, Yifan Yang, Qingyuan Wang, Juntuo Wang, Yuwei Zhang, Ru Zhen, Yanhao Zhang, Haonan Lu, Xudong Zhang

AAAI 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 summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Targets cross-view spatial reasoning in LVLMs, where prior work only handles single-frame 3D understanding; the abstract is truncated so the proposed method is not stated.

Problem

LVLMs struggle with consistent cross-view spatial reasoning, since existing methods inject 3D spatial concepts under a single coordinate system within one frame.

Approach

Abstract is truncated; the paper is said to align cross-view visual geometries via human-like reasoning learning, but the specific mechanism is not stated.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add human-like reasoning data (ablation)SFT+RL CVVG-Reasoner · MV3DSR-Benchcamera pose: 0.384 -> 0.69Improved
add human-like reasoning data (ablation)SFT+RL CVVG-Reasoner · MV3DSR-Benchmotion tracking: 0.571 -> 0.773Improved
add human-like reasoning data (ablation)SFT+RL CVVG-Reasoner · MV3DSR-Benchpose-aware shifts: 0.512 -> 0.588Improved
add human-like reasoning data (ablation)SFT+RL CVVG-Reasoner · MV3DSR-Benchcross-view grounding: 0.46 -> 0.45No clear effect
add RL stage after SFT (ablation)CVVG-Reasoner · MV3DSR-Benchcamera pose: 0.392 -> 0.69Improved
add RL stage after SFT (ablation)CVVG-Reasoner · MV3DSR-Benchmotion tracking: 0.484 -> 0.773Improved
add RL stage after SFT (ablation)CVVG-Reasoner · MV3DSR-Benchpose-aware shifts: 0.385 -> 0.588Improved
add RL stage after SFT (ablation)CVVG-Reasoner · MV3DSR-Benchcross-view grounding: 0.331 -> 0.45Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Limitations

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