AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Imaginative Perception Tokens Enhance Spatial Reasoning in Multimodal Language Models

Mahtab Bigverdi, Linjie Li, Weikai Huang, Jieyu Zhang, Tuhin Kundu, Zelun Luo, Chris Kim, Jaemin Cho, Ranjay Krishna, Linda Shapiro, Liu Yiming

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
answer-only fine-tuning on AI2-THOR dataBagel (base) · AI2-THOR PETaccuracy (%): 40.3 -> 97.5Improved
add IPT supervision vs label-only (ablation)Bagel · AI2-THOR MVCaccuracy (%): 63.9 -> 67.3Improved
add IPT supervision vs label-only (ablation)Bagel · Habitat PETaccuracy (%): 82.0 -> 87.0Improved
add IPT supervision vs label-only (ablation)Bagel · AI2-THOR PT (avg)accuracy (%): 65.7 -> 49.0Hurt
add IPT supervision vs label-only (ablation)Bagel · AI2-THOR PETaccuracy (%): 97.5 -> 96.8No clear effect
add textual chain-of-thought vs label-only (ablation)Bagel · AI2-THOR PETaccuracy (%): 97.5 -> 83.1Hurt
add textual chain-of-thought vs label-only (ablation)Bagel · AI2-THOR PT (avg)accuracy (%): 65.7 -> 49.7Hurt
replace Text CoT with IPT (ablation)Bagel · AI2-THOR PETaccuracy (%): 83.1 -> 96.8Improved
mix IPT and label-only data vs label-only (ablation)Bagel · Real PT (different env)accuracy (%): 54.7 -> 58.6Improved
mix IPT and label-only data vs IPT (ablation)Bagel · AI2-THOR MVCaccuracy (%): 67.3 -> 62.3Hurt
raise imagination resolution Latent-4 to Latent-64 (ablation)Bagel + IPT · AI2-THOR MVCaccuracy (%): 53.5 -> 63.1Improved
raise imagination resolution Latent-32 to Latent-64 (ablation)Bagel + IPT · Habitat PETaccuracy (%): 87.0 -> 83.3Hurt

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