aiqbench analysis
Imaginative Perception Tokens (IPT) offer a principled approach to enhancing spatial reasoning in vision language models by externalizing unobserved spatial configurations without requiring image generation at inference time. This method outperforms textual chain-of-thought training on specific benchmarks, addressing a critical modality mismatch where forcing spatial computation through language degrades performance. The technique provides interpretable intermediate representations that improve generalization on tasks like perspective taking and path tracing, marking a significant step toward robust multimodal reasoning.
Imaginative Perception Tokens Enhance Spatial Reasoning in Multimodal Language Models
arXiv:2606.03988v2 Announce Type: replace Abstract: Vision language models (VLMs) excel at many tasks but still struggle with spatial reasoning when critical information is not directly observable. Many such problems require imaginative perception: inferring what would be seen from an unseen viewpoint, tracing paths through occluded spaces, or integrating partial observations into a coherent spatial representation. We introduce Imaginative Perception Tokens (IPT), intermediate perceptual representations that externalize what a VLM would perceive under alternative spatial configurations while remaining consistent with the observed input. To study this capability, we formulate three tasks, Perspective Taking (PET), Path Tracing (PT), and Multiview Counting (MVC), and construct datasets of approximately 20K examples with ground truth imaginations, answers, and evaluation benchmarks. Using the unified VLM BAGEL as the backbone, IPT supervision consistently improves spatial reasoning and often outperforms textual chain of thought training, even without generating images at inference time. On MVC, IPT improves accuracy by 3.4% and achieves competitive performance with strong closed-source models on PT. We further find that combining IPT and label-only supervision yields additional gains, whereas textual chain of thought can substantially degrade performance, suggesting a modality mismatch when spatial computation is forced through language. Overall, IPT provides a principled supervision signal for reasoning about unobserved spatial structure, improving generalization while producing interpretable intermediate representations.
Abstract:Vision language models (VLMs) excel at many tasks but still struggle with spatial reasoning when critical information is not directly observable. Many such problems require imaginative perception: inferring what would be seen from an unseen viewpoint, tracing paths through occluded spaces, or integrating partial observations into a coherent spatial representation. We introduce Imaginative Perception Tokens (IPT), intermediate perceptual representations that externalize what a VLM would perceive under alternative spatial configurations while remaining consistent with the observed input. To study this capability, we formulate three tasks, Perspective Taking (PET), Path Tracing (PT), and Multiview Counting (MVC), and construct datasets of approximately 20K examples with ground truth imaginations, answers, and evaluation benchmarks. Using the unified VLM BAGEL as the backbone, IPT supervision consistently improves spatial reasoning and often outperforms textual chain of thought training, even without generating images at inference time. On MVC, IPT improves accuracy by 3.4% and achieves competitive performance with strong closed-source models on PT. We further find that combining IPT and label-only supervision yields additional gains, whereas textual chain of thought can substantially degrade performance, suggesting a modality mismatch when spatial computation is forced through language. Overall, IPT provides a principled supervision signal for reasoning about unobserved spatial structure, improving generalization while producing interpretable intermediate representations.
Artificial Intelligence (cs.AI)
arXiv:2606.03988 [cs.AI]
(or arXiv:2606.03988v2 [cs.AI] for this version)
Full text under the publisher's open licence · 219 words
Related coverage
Imaginative Perception Tokens Enhance Spatial Reasoning in Multimodal Language Models
arXiv · Aug 18, 2026
Imaginative Perception Tokens Enhance Spatial Reasoning in Multimodal Language Models
arXiv · Jun 3, 2026
Scaling Native Multimodal Pre-Training From Scratch
arXiv · Jul 27, 2026
GeoAnchor: Collaborative Reasoning via Latent Decomposition for 3D Spatial Understanding
arXiv · Jul 29, 2026