# GLIMPSE: Holistic Cross-Modal Explainability for Large Vision-Language Models

Source: [GLIMPSE: Holistic Cross-Modal Explainability for Large Vision-Language Models](<https://www.sota2.com/research/paper/68fa7a4d1e209131cea824f2>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-26T05:17:11.038Z

Authors: Guanxi Shen

Published: 2025-06-23T18:00:04.000Z

## Abstract

Recent large vision-language models (LVLMs) have advanced capabilities in visual question answering (VQA). However, interpreting where LVLMs direct their visual attention remains a significant challenge, yet is essential for understanding model behavior. We introduce GLIMPSE (Gradient-Layer Importance Mapping for Prompted Visual Saliency Explanation), a lightweight, model-agnostic framework that jointly attributes LVLM outputs to the most relevant visual evidence and textual signals that support open-ended generation. GLIMPSE fuses gradient-weighted attention, adaptive layer propagation, and relevance-weighted token aggregation to produce holistic response-level heat maps for interpreting cross-modal reasoning, outperforming prior methods in faithfulness and pushing the state-of-the-art in human-attention alignment. We demonstrate an analytic approach to uncover fine-grained insights into LVLM cross-modal attribution, trace reasoning dynamics, analyze systematic misalignment, diagnose hallucination and bias, and ensure transparency.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2506.18985>)

[PDF](<https://arxiv.org/pdf/2506.18985.pdf>)
