Token Warping Helps MLLMs Look from Nearby Viewpoints

arXiv cs.CV / 4/6/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper investigates whether warping image tokens (not pixels) can help multimodal large language models (MLLMs) understand scenes from nearby viewpoints, addressing the fragility of pixel-wise warping to depth errors.
It compares forward vs. backward token warping and finds backward token warping is more stable and better preserves semantic coherence during viewpoint shifts.
The authors introduce and evaluate on the ViewBench benchmark, showing that token-level warping enables more reliable viewpoint reasoning.
Results indicate token-level warping outperforms multiple baselines, including pixel-wise warping, spatially fine-tuned MLLMs, and a generative warping method.

Abstract

Can warping tokens, rather than pixels, help multimodal large language models (MLLMs) understand how a scene appears from a nearby viewpoint? While MLLMs perform well on visual reasoning, they remain fragile to viewpoint changes, as pixel-wise warping is highly sensitive to small depth errors and often introduces geometric distortions. Drawing on theories of mental imagery that posit part-level structural representations as the basis for human perspective transformation, we examine whether image tokens in ViT-based MLLMs serve as an effective substrate for viewpoint changes. We compare forward and backward warping, finding that backward token warping, which defines a dense grid on the target view and retrieves a corresponding source-view token for each grid point, achieves greater stability and better preserves semantic coherence under viewpoint shifts. Experiments on our proposed ViewBench benchmark demonstrate that token-level warping enables MLLMs to reason reliably from nearby viewpoints, consistently outperforming all baselines including pixel-wise warping approaches, spatially fine-tuned MLLMs, and a generative warping method.