Cover Image

Why Is Spatial Reasoning Hard for VLMs? An Attention Mechanism Perspective on Focus Areas

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Chen, Shiqi, Zhu, Tongyao, Zhou, Ruochen, Zhang, Jinghan, Gao, Siyang, Niebles, Juan Carlos, Geva, Mor, He, Junxian, Wu, Jiajun, Li, Manling
Format:	Preprint
Published:	2025
Subjects:	Computation and Language
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Bring Reason to Vision: Understanding Perception and Reasoning through Model Merging
by: Chen, Shiqi, et al.
Published: (2025)

Internalizing World Models via Self-Play Finetuning for Agentic RL
by: Chen, Shiqi, et al.
Published: (2025)

SkillCraft: Can LLM Agents Learn to Use Tools Skillfully?
by: Chen, Shiqi, et al.
Published: (2026)

Linear Scaling Video VLMs for Long Video Understanding
by: Eyzaguirre, Cristobal, et al.
Published: (2026)

Inferring Functionality of Attention Heads from their Parameters
by: Elhelo, Amit, et al.
Published: (2024)

In-Context Sharpness as Alerts: An Inner Representation Perspective for Hallucination Mitigation
by: Chen, Shiqi, et al.
Published: (2024)

Focusing by Contrastive Attention: Enhancing VLMs' Visual Reasoning
by: Ge, Yuyao, et al.
Published: (2025)

Universal Jailbreak Suffixes Are Strong Attention Hijackers
by: Ben-Tov, Matan, et al.
Published: (2025)

Estimating Knowledge in Large Language Models Without Generating a Single Token
by: Gottesman, Daniela, et al.
Published: (2024)

Mixing Mechanisms: How Language Models Retrieve Bound Entities In-Context
by: Gur-Arieh, Yoav, et al.
Published: (2025)

Constructing Interpretable Features from Compositional Neuron Groups
by: Shafran, Or, et al.
Published: (2025)

Language Models Encode Numbers Using Digit Representations in Base 10
by: Levy, Amit Arnold, et al.
Published: (2024)

Does Learning Mathematical Problem-Solving Generalize to Broader Reasoning?
by: Zhou, Ruochen, et al.
Published: (2025)

Data-driven Circuit Discovery for Interpretability of Language Models
by: Rai, Daking, et al.
Published: (2026)

Eliciting Textual Descriptions from Representations of Continuous Prompts
by: Ramati, Dana, et al.
Published: (2024)

Narrowing the Knowledge Evaluation Gap: Open-Domain Question Answering with Multi-Granularity Answers
by: Yona, Gal, et al.
Published: (2024)

Preventing Rogue Agents Improves Multi-Agent Collaboration
by: Barbi, Ohav, et al.
Published: (2025)

Can Large Language Models Faithfully Express Their Intrinsic Uncertainty in Words?
by: Yona, Gal, et al.
Published: (2024)

Hallucinations Undermine Trust; Metacognition is a Way Forward
by: Yona, Gal, et al.
Published: (2026)

Routers Learn the Geometry of Their Experts: Geometric Coupling in Sparse Mixture-of-Experts
by: Ahrac, Sagi, et al.
Published: (2026)

SpinBench: Perspective and Rotation as a Lens on Spatial Reasoning in VLMs
by: Zhang, Yuyou, et al.
Published: (2025)

Do Large Language Models Latently Perform Multi-Hop Reasoning?
by: Yang, Sohee, et al.
Published: (2024)

Fine-Grained Preference Optimization Improves Spatial Reasoning in VLMs
by: Shen, Yifan, et al.
Published: (2025)

Do Large Language Models Perform Latent Multi-Hop Reasoning without Exploiting Shortcuts?
by: Yang, Sohee, et al.
Published: (2024)

Adaptive Chain-of-Focus Reasoning via Dynamic Visual Search and Zooming for Efficient VLMs
by: Zhang, Xintong, et al.
Published: (2025)

Faithfulness Metrics Don't Measure Faithfulness: A Meta-Evaluation with Ground Truth
by: Gur-Arieh, Yoav, et al.
Published: (2026)

Mechanistically Interpretable Neural Encoding Reveals Fine-Grained Functional Selectivity in Human Visual Cortex
by: Grosbard, Idan Daniel, et al.
Published: (2026)

Disentangling MLP Neuron Weights in Vocabulary Space
by: Avrahamy, Asaf, et al.
Published: (2026)

Detecting (Un)answerability in Large Language Models with Linear Directions
by: Lavi, Maor Juliet, et al.
Published: (2025)

Compression Represents Intelligence Linearly
by: Huang, Yuzhen, et al.
Published: (2024)

On the Perception Bottleneck of VLMs for Chart Understanding
by: Liu, Junteng, et al.
Published: (2025)

World2VLM: Distilling World Model Imagination into VLMs for Dynamic Spatial Reasoning
by: Zhang, Wanyue, et al.
Published: (2026)

Thinking to Recall: How Reasoning Unlocks Parametric Knowledge in LLMs
by: Gekhman, Zorik, et al.
Published: (2026)

Focus Directions Make Your Language Models Pay More Attention to Relevant Contexts
by: Zhu, Youxiang, et al.
Published: (2025)

Reasoning Path and Latent State Analysis for Multi-view Visual Spatial Reasoning: A Cognitive Science Perspective
by: Xue, Qiyao, et al.
Published: (2025)

Image-of-Thought Prompting for Visual Reasoning Refinement in Multimodal Large Language Models
by: Zhou, Qiji, et al.
Published: (2024)

Reasmory: 3D Reconstruction as Explicit Memory for VLMs Spatial Reasoning
by: He, Jixuan, et al.
Published: (2026)

The Gordian Knot for VLMs: Diagrammatic Knot Reasoning as a Hard Benchmark
by: Liu, Hao, et al.
Published: (2026)

Performance Gap in Entity Knowledge Extraction Across Modalities in Vision Language Models
by: Cohen, Ido, et al.
Published: (2024)

Backward Lens: Projecting Language Model Gradients into the Vocabulary Space
by: Katz, Shahar, et al.
Published: (2024)