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

• URL: http://arxiv.org/abs/2509.14837v1
• Date: Thu, 18 Sep 2025 10:58:34 GMT
• Title: V-SEAM: Visual Semantic Editing and Attention Modulating for Causal Interpretability of Vision-Language Models
• Authors: Qidong Wang, Junjie Hu, Ming Jiang,
• Abstract summary: We introduce V-SEAM, a novel framework that combines Visual Semantic Editing and Attention Modulating for causal interpretation of vision-language models.<n>V-SEAM identifies attention heads with positive or negative contributions to predictions across three semantic levels.<n>We demonstrate enhanced performance for both LLaVA and InstructBLIP across three diverse VQA benchmarks.
• Score: 10.052877942432783
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Recent advances in causal interpretability have extended from language models to vision-language models (VLMs), seeking to reveal their internal mechanisms through input interventions. While textual interventions often target semantics, visual interventions typically rely on coarse pixel-level perturbations, limiting semantic insights on multimodal integration. In this study, we introduce V-SEAM, a novel framework that combines Visual Semantic Editing and Attention Modulating for causal interpretation of VLMs. V-SEAM enables concept-level visual manipulations and identifies attention heads with positive or negative contributions to predictions across three semantic levels: objects, attributes, and relationships. We observe that positive heads are often shared within the same semantic level but vary across levels, while negative heads tend to generalize broadly. Finally, we introduce an automatic method to modulate key head embeddings, demonstrating enhanced performance for both LLaVA and InstructBLIP across three diverse VQA benchmarks. Our data and code are released at: https://github.com/petergit1/V-SEAM.

Related papers

• Youtu-VL: Unleashing Visual Potential via Unified Vision-Language Supervision [79.06371915084833]
  We introduce Youtu-VL, a framework leveraging the Vision-Language Unified Autoregressive Supervision (VLUAS) paradigm.<n>Youtu-VL applies unified autoregressive supervision to both visual details and linguistic content.<n>We extend this paradigm to encompass vision-centric tasks, enabling a standard VLM to perform vision-centric tasks without task-specific additions.
  arXiv  Detail & Related papers  (2026-01-27T17:01:16Z)
• AMVICC: A Novel Benchmark for Cross-Modal Failure Mode Profiling for VLMs and IGMs [2.357397994148727]
  multimodal large language models (MLLMs) and image generation models (IGMs) were investigated.<n>We created a novel benchmark to compare failure modes across image-to-text and text-to-image tasks.<n>Our results show that failure modes are often shared between models and modalities, but certain failures are model-specific and modality-specific.
  arXiv  Detail & Related papers  (2026-01-20T00:06:58Z)
• Attention Guided Alignment in Efficient Vision-Language Models [56.20286899428444]
  Large Vision-Language Models (VLMs) rely on effective multimodal alignment between pre-trained vision encoders and Large Language Models (LLMs)<n>This paper presents a comprehensive analysis of attention patterns in efficient VLMs.<n>We introduce Attention-Guided Efficient Vision-Language Models (AGE-VLM), a novel framework that enhances visual grounding through interleaved cross-attention layers.
  arXiv  Detail & Related papers  (2025-11-21T21:36:48Z)
• When Alignment Fails: Multimodal Adversarial Attacks on Vision-Language-Action Models [75.16145284285456]
  We introduce VLA-Fool, a comprehensive study of multimodal adversarial robustness in embodied VLA models under both white-box and black-box settings.<n>We develop the first automatically crafted and semantically guided prompting framework.<n> Experiments on the LIBERO benchmark reveal that even minor multimodal perturbations can cause significant behavioral deviations.
  arXiv  Detail & Related papers  (2025-11-20T10:14:32Z)
• VL-SAE: Interpreting and Enhancing Vision-Language Alignment with a Unified Concept Set [80.50996301430108]
  The alignment of vision-language representations endows current Vision-Language Models with strong multi-modal reasoning capabilities.<n>We propose VL-SAE, a sparse autoencoder that encodes vision-language representations into its hidden activations.<n>For interpretation, the alignment between vision and language representations can be understood by comparing their semantics with concepts.
  arXiv  Detail & Related papers  (2025-10-24T10:29:31Z)
• Knowledge Transfer from Interaction Learning [16.83991179271622]
  Current visual foundation models (VFMs) face a fundamental limitation in transferring knowledge from vision language models (VLMs)<n>We propose Learning from Interactions (LFI), a cognitive-inspired framework that explicitly modeling visual understanding as an interactive process.
  arXiv  Detail & Related papers  (2025-09-23T07:27:36Z)
• SECOND: Mitigating Perceptual Hallucination in Vision-Language Models via Selective and Contrastive Decoding [5.976839106353883]
  SECOND: Selective and Contrastive Decoding is a novel approach that enables Vision-Language Models to leverage multi-scale visual information with an object-centric manner.<n> SECOND significantly reduces perceptual hallucinations and outperforms a wide range of benchmarks.
  arXiv  Detail & Related papers  (2025-06-10T02:55:38Z)
• Exploring Implicit Visual Misunderstandings in Multimodal Large Language Models through Attention Analysis [21.869968563545736]
  We define implicit visual misunderstanding (IVM), where MLLMs provide correct answers without fully comprehending the visual input.<n>We introduce a scale-agnostic metric, textitattention accuracy, and a novel benchmark for quantifying IVMs.<n>We extend our approach to finer granularities and demonstrate its effectiveness in unimodal scenarios.
  arXiv  Detail & Related papers  (2025-05-15T17:52:40Z)
• Text Speaks Louder than Vision: ASCII Art Reveals Textual Biases in Vision-Language Models [93.46875303598577]
  Vision-language models (VLMs) have advanced rapidly in processing multimodal information, but their ability to reconcile conflicting signals remains underexplored.<n>This work investigates how VLMs process ASCII art, a unique medium where textual elements collectively form visual patterns, potentially creating semantic-visual conflicts.
  arXiv  Detail & Related papers  (2025-04-02T10:47:07Z)
• Prismatic VLMs: Investigating the Design Space of Visually-Conditioned Language Models [73.40350756742231]
  Visually-conditioned language models (VLMs) have seen growing adoption in applications such as visual dialogue, scene understanding, and robotic task planning. Despite the volume of new releases, key design decisions around image preprocessing, architecture, and optimization are under-explored.
  arXiv  Detail & Related papers  (2024-02-12T18:21:14Z)
• DiMBERT: Learning Vision-Language Grounded Representations with Disentangled Multimodal-Attention [101.99313208598569]
  Vision-and-language (V-L) tasks require the system to understand both vision content and natural language. We propose DiMBERT (short for Disentangled Multimodal-Attention BERT), which applies separated attention spaces for vision and language. We show that DiMBERT sets new state-of-the-art performance on three tasks.
  arXiv  Detail & Related papers  (2022-10-28T23:00:40Z)
• Linguistically-aware Attention for Reducing the Semantic-Gap in Vision-Language Tasks [9.462808515258464]
  We propose an attention mechanism - Linguistically-aware Attention (LAT) - that leverages object attributes obtained from generic object detectors. LAT represents visual and textual modalities in a common linguistically-rich space, thus providing linguistic awareness to the attention process. We apply and demonstrate the effectiveness of LAT in three Vision-language (V-L) tasks: Counting-VQA, VQA, and Image captioning.
  arXiv  Detail & Related papers  (2020-08-18T16:29:49Z)
• Behind the Scene: Revealing the Secrets of Pre-trained Vision-and-Language Models [65.19308052012858]
  Recent Transformer-based large-scale pre-trained models have revolutionized vision-and-language (V+L) research. We present VALUE, a set of meticulously designed probing tasks to decipher the inner workings of multimodal pre-training. Key observations: Pre-trained models exhibit a propensity for attending over text rather than images during inference.
  arXiv  Detail & Related papers  (2020-05-15T01:06:54Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.