AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

VSP: Diagnosing the Dual Challenges of Perception and Reasoning in Spatial Planning Tasks for MLLMs

Qiucheng Wu, Handong Zhao, Michael Saxon, Trung Bui, William Yang Wang, Yang Zhang, Shiyu Chang

ICCV 2025 · Vision-language and multimodal › Multimodal LLMs · approach: Spatial reasoning in VLMs · 1 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
LoRA fine-tune on 10k VSP examplesLLaVA-v1.6-Vicuna-7B · Maze Navigation main tasksuccess rate: 0.05 -> 0.60Improved
LoRA fine-tune on 10k VSP examplesLLaVA-v1.6-Vicuna-7B · Blocks World main tasksuccess rate: 0.01 -> 0.97Improved
LoRA fine-tune on 10k VSP examplesInternLM-XComposer2-7B · Maze Navigation main tasksuccess rate: 0.11 -> 0.17Improved
LoRA fine-tune on 10k VSP examplesInternLM-XComposer2-7B · Blocks World main tasksuccess rate: 0.00 -> 0.09Improved
add 4 in-context examplesGPT-4V · Maze Navigation T2accuracy: 0.27 -> 0.69Improved
add 2 in-context examplesGemini-1.0 · Blocks World T3accuracy: 0.54 -> 0.76Improved
add 4 in-context examplesGemini-1.0 · Maze Navigation main tasksuccess rate: 0.17 -> 0.19No clear effect
add 4 in-context examplesGPT-4V · Maze Navigation T1accuracy: 0.56 -> 0.54No clear effect
add 4 in-context examplesGemini-1.0 · Blocks World main tasksuccess rate: 0.03 -> 0.06No clear effect
increase maze size from 3x3 to 8x8GPT-4V · Maze Navigation main tasksuccess rate: 0.55 -> 0.10Hurt
replace image input with textual description of environment (ablation)Gemini · Maze Navigation and Blocks World tasksaccuracyImproved
upgrade Gemini-1.0 to Gemini-2.0Gemini · VSP overallsuccess rate: 0.1167 -> 0.4533Improved

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