AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

GeoMIND: A Benchmark for Spatial Understanding in Robotic Manipulation

Jinghe Wang, Xinrui Cao, Duo Wu, Chenghao Gu, Yong Zhong, Linjia Kang, Tianyi Xiong, Zhi Wang

NeurIPS 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.

Introduces GeoMIND, a benchmark for evaluating spatial understanding (spatial reasoning and spatial memory) in robotic manipulation, per the abstract.

Problem

Robotic manipulation in open-world settings needs robust spatial understanding (identifying targets in complex layouts and adapting to scenes that evolve through interaction), but existing embodied benchmarks are not stated in the abstract beyond being insufficient.

Approach

A benchmark framing spatial understanding for manipulation as two capabilities: spatial reasoning and spatial memory. Details beyond the abstract are not available.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
specify target by name and visual cue (ablation)pi0.5 · GeoMIND-Basesuccess rate (%): 11.3 -> 39.2Improved
add semantic augmentation to instructions (ablation)six VLA policies · GeoMINDsuccess rate (%): 22.6 -> 14.2 (avg)Hurt
add visual augmentation (ablation)OpenVLA · GeoMINDsuccess rate (%): 25.0 -> 12.5Hurt
add visual augmentation (ablation)pi0 · GeoMINDsuccess rate (%): 16.7 -> 16.7No clear effect
add visual augmentation (ablation)InternVLA-M1 · GeoMINDsuccess rate (%): 37.5 -> 37.5No clear effect
add execution augmentation (ablation)pi0 · GeoMINDsuccess rate (%): 16.7 -> 10.2Hurt
add execution augmentation (ablation)InternVLA-M1 · GeoMINDsuccess rate (%): 37.5 -> 35.4Hurt
add spatial enhancement (InternVLA-M1 vs general policies)VLA policies · GeoMIND-Basesuccess rate (%): 26.4 vs 20.7 (OpenVLA)Mixed
use modular VLM grounding with scripted execution instead of end-to-endCOPA/MOKA · GeoMIND-Basesuccess rate (%): 4.0-16.9No clear effect
use Claude-Sonnet-4.6 backbone instead of Gemini-2.5-FlashCOPA · GeoMIND-Basesuccess rate (%): 4.6 -> 15.9Improved

Experimental setup

Benchmarks: GeoMIND

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