AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

GRASP: Awakening Latent Spatial Reasoning in LVLMs via Training-free Geometric Rectification

Jiadong Yan, Ke Zhang, Chenyang Zhao, Shoushan Li, Xizhao Luo

ICML 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Spatial reasoning in VLMs

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.

Shows via linear probing that LVLMs encode spatial features internally yet output wrong answers, and proposes GRASP, a training-free inference-time geometric rectification method to close this gap.

Problem

LVLMs struggle with spatial reasoning despite internally encoding spatial features; the internal representation and final text output are misaligned.

Approach

Inference-time geometric manifold adaptation that rectifies internal representations using geometric structure, without any weight updates, to surface latent spatial knowledge in the output.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add Implicit Trajectory Correction (ablation)LLaVA-1.5 7B · WhatsUp Controlled Aaccuracy: 60.3 -> 83.9Improved
add Explicit Distribution Alignment (ablation)LLaVA-1.5 7B · WhatsUp Controlled Aaccuracy: 60.3 -> 73.3Improved
add EDA to ITC (full GRASP) (ablation)LLaVA-1.5 7B · WhatsUp Controlled Aaccuracy: 83.9 -> 86.4Improved
add ITC to EDA (full GRASP) (ablation)Qwen2.5-VL 3B · WhatsUp Controlled Aaccuracy: 95.8 -> 97.6Improved
add ITC and EDA to the base model (ablation)LLaVA-NeXT-Video 7B · VSI-Bench Relative Direction (Hard)accuracy: 31.1 -> 39.0Improved
add ITC and EDA to the base model (ablation)Qwen3-VL 8B · VSI-Bench Relative Direction (Hard)accuracy: 35.9 -> 45.6Improved
inject correction into self-attention instead of FFN (ablation)ITC · WhatsUp Controlled A (LLaVA-1.5)accuracy: 69.7 -> 83.9Improved
inject correction into self-attention instead of FFN (ablation)ITC · WhatsUp Controlled A (Qwen2.5-VL)accuracy: 90.6 -> 93.3Improved
increase manifold search candidates N from 1 to 2 (ablation)GRASP · WhatsUp Controlled A (LLaVA-1.5)accuracy: 81.6 -> 86.4Improved
increase manifold search candidates N from 2 to 4 (ablation)GRASP · WhatsUp Controlled A (LLaVA-1.5)accuracy: 86.4 -> 86.4No clear effect
increase optimization steps T from 1 to 2 (ablation)GRASP · WhatsUp Controlled A (LLaVA-1.5)accuracy: 84.8 -> 86.4Improved
increase search radius from 5 to 10 degrees (ablation)GRASP · WhatsUp Controlled A (LLaVA-1.5)accuracy: 78.8 -> 85.9Improved

Surprising findings

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