Pith. sign in

REVIEW 6 cited by

Where do Large Vision-Language Models Look at when Answering Questions?

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2503.13891 v1 pith:VXHOJU22 submitted 2025-03-18 cs.CV cs.CL

Where do Large Vision-Language Models Look at when Answering Questions?

classification cs.CV cs.CL
keywords visuallvlmsunderstandingvision-languageansweransweringimageinput
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Large Vision-Language Models (LVLMs) have shown promising performance in vision-language understanding and reasoning tasks. However, their visual understanding behaviors remain underexplored. A fundamental question arises: to what extent do LVLMs rely on visual input, and which image regions contribute to their responses? It is non-trivial to interpret the free-form generation of LVLMs due to their complicated visual architecture (e.g., multiple encoders and multi-resolution) and variable-length outputs. In this paper, we extend existing heatmap visualization methods (e.g., iGOS++) to support LVLMs for open-ended visual question answering. We propose a method to select visually relevant tokens that reflect the relevance between generated answers and input image. Furthermore, we conduct a comprehensive analysis of state-of-the-art LVLMs on benchmarks designed to require visual information to answer. Our findings offer several insights into LVLM behavior, including the relationship between focus region and answer correctness, differences in visual attention across architectures, and the impact of LLM scale on visual understanding. The code and data are available at https://github.com/bytedance/LVLM_Interpretation.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 6 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Measuring Cross-Modal Synergy: A Benchmark for VLM Explainability

    cs.AI 2026-05 unverdicted novelty 7.0

    Introduces Synergistic Faithfulness metric based on Shapley Interaction Index to evaluate cross-modal synergy in VLM explainers, revealing over-reliance on visual salience in existing methods.

  2. V-SEAM: Visual Semantic Editing and Attention Modulating for Causal Interpretability of Vision-Language Models

    cs.CL 2025-09 conditional novelty 7.0

    V-SEAM combines concept-level visual semantic editing with attention head modulation to identify positive and negative contributors across object, attribute, and relationship levels, then uses this to improve VLM perf...

  3. PhaseWin: An Efficient Search Algorithm for Faithful Visual Attribution

    cs.CV 2026-06 unverdicted novelty 6.0

    PhaseWin is a phased window-search algorithm for faithful visual attribution that achieves linear evaluation complexity with near-greedy faithfulness under monotone evidence-accumulation and feature-level structural a...

  4. Where Not to Learn: Prior-Aligned Training with Subset-based Attribution Constraints for Reliable Decision-Making

    cs.CV 2026-01 unverdicted novelty 5.0

    A training approach that adds attribution constraints derived from human priors to steer models toward using specified input regions for decisions.

  5. Where Not to Learn: Prior-Aligned Training with Subset-based Attribution Constraints for Reliable Decision-Making

    cs.CV 2026-01 conditional novelty 5.0

    Penalizing off-prior attribution evidence during training with subset-selection attribution improves accuracy and attribution reasonability in image classifiers and MLLM-based GUI agents.

  6. LaSM: Layer-wise Scaling Mechanism for Defending Pop-up Attack on GUI Agents

    cs.CR 2025-07 conditional novelty 5.0

    LaSM is a layer-wise scaling mechanism that amplifies attention and MLP modules in critical layers to defend GUI agents against pop-up attacks by correcting attention misalignment.