LISA: A Layer-wise Integration and Suppression Approach for Hallucination Mitigation in Multimodal Large Language Models

• URL: http://arxiv.org/abs/2507.19110v1
• Date: Fri, 25 Jul 2025 09:48:23 GMT
• Title: LISA: A Layer-wise Integration and Suppression Approach for Hallucination Mitigation in Multimodal Large Language Models
• Authors: Zhihui Guo, Xin Man, Hui Xu, Jie Shao,
• Abstract summary: Multimodal Large Language Models (MLLMs) excel in vision-language tasks but remain prone to object hallucinations.<n>We propose textbfLISA, which enhances generation consistency through hierarchical modulation and multi-layer fusion.<n>Experiments show that LISA reduces hallucinations by up to 53.6% in $mathrmCHAIR_I$ and improves POPE F1 by 4.5%.
• Score: 8.122679857175315
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Multimodal Large Language Models (MLLMs) excel in vision-language tasks such as image captioning but remain prone to object hallucinations, where they describe objects that do not appear in the image. To mitigate this, we propose \textbf{LISA}, a \textbf{L}ayer-wise \textbf{I}ntegration and \textbf{S}uppression \textbf{A}pproach that enhances generation consistency through hierarchical modulation and multi-layer fusion. LISA leverages the functional hierarchy within MLLMs, where shallow layers provide visual grounding, middle layers encode semantics, and deep layers tend to amplify spurious signals. First, zone-specific spectral modulation stabilizes attention by suppressing over-amplified activations in deeper layers while preserving alignment cues in earlier layers. Second, token-level logits from selected layers are fused via anchor-based routing, with token-wise anchor selection and soft logit fusion enabling adaptive integration during decoding. LISA is fully \textbf{plug-and-play} and can be seamlessly integrated into existing MLLMs, including Qwen2.5-VL. Experiments on multiple benchmarks show that LISA reduces hallucinations by up to 53.6\% in $\mathrm{CHAIR}_I$ and improves POPE F1 by 4.5\%, demonstrating strong generalization across models and tasks.

Related papers

• From One-to-One to Many-to-Many: Dynamic Cross-Layer Injection for Deep Vision-Language Fusion [91.35078719566472]
  Vision-Language Models (VLMs) create a severe visual feature bottleneck by using a crude, asymmetric connection.<n>We introduce Cross-Layer Injection (CLI), a novel and lightweight framework that forges a dynamic many-to-many bridge between the two modalities.
  arXiv  Detail & Related papers  (2026-01-15T18:59:10Z)
• PlaM: Training-Free Plateau-Guided Model Merging for Better Visual Grounding in MLLMs [59.78917775399492]
  Multimodal instruction fine-tuning paradoxically degrades this text's reasoning capability.<n>We propose a training-free framework to mitigate this degradation.
  arXiv  Detail & Related papers  (2026-01-12T15:27:51Z)
• Text-Guided Layer Fusion Mitigates Hallucination in Multimodal LLMs [25.843085393058434]
  TGIF (Text-Guided Inter-layer Fusion) is a lightweight module that treats encoder layers as depth-wise "experts"<n> TGIF provides consistent improvements across hallucination, OCR, and VQA benchmarks.
  arXiv  Detail & Related papers  (2026-01-06T15:31:19Z)
• A Comprehensive Study on Visual Token Redundancy for Discrete Diffusion-based Multimodal Large Language Models [85.30893355216486]
  We study how visual token redundancy evolves with different dMLLM architectures and tasks.<n>Our study reveals that visual redundancy emerges only in from-scratch dMLLMs while handling long-answer tasks.<n>Layer-skipping is promising for accelerating AR-to-diffusion dMLLMs, whereas progressive or late-step pruning is more effective for from-scratch dMLLMs.
  arXiv  Detail & Related papers  (2025-11-19T04:13:36Z)
• Towards Fine-Grained Vision-Language Alignment for Few-Shot Anomaly Detection [65.29550320117526]
  We propose a novel framework named FineGrainedAD to improve anomaly localization performance.<n> Experiments demonstrate that the proposed FineGrainedAD achieves superior overall performance in few-shot settings.
  arXiv  Detail & Related papers  (2025-10-30T13:09:00Z)
• D$^2$HScore: Reasoning-Aware Hallucination Detection via Semantic Breadth and Depth Analysis in LLMs [15.665202830841046]
  This work revisits hallucination detection from the perspective of model architecture and generation dynamics.<n>We propose textbfD$2$HScore (Dispersion and Drift-based Hallucination Score), a training-free and label-free framework.<n>Experiments across five open-source Language Models and five widely used benchmarks demonstrate that D$2$HScore consistently outperforms existing training-free baselines.
  arXiv  Detail & Related papers  (2025-09-15T04:28:38Z)
• Dynamic Embedding of Hierarchical Visual Features for Efficient Vision-Language Fine-Tuning [5.85033069870214]
  We propose an efficient vision-language fine-tuning method based on dynamic embedding and fusion of hierarchical visual features.<n>By fine-tuning only a small number of parameters, DEHVF achieves precise alignment and complement of cross-modal information.
  arXiv  Detail & Related papers  (2025-08-25T03:57:46Z)
• ByDeWay: Boost Your multimodal LLM with DEpth prompting in a Training-Free Way [7.18701660596182]
  ByDeWay is a training-free framework designed to enhance the performance of Multimodal Large Language Models (MLLMs)<n>ByDeWay uses a novel prompting strategy called Layered-Depth-Based Prompting (LDP)<n>It segments the scene into closest, mid-range, and farthest layers using monocular depth estimation, then generates region-specific captions with a grounded vision-language model.
  arXiv  Detail & Related papers  (2025-07-11T15:21:49Z)
• Rethinking Visual Layer Selection in Multimodal LLMs [46.091556112958884]
  This work proposes a Layer-wise Similarity approach to group CLIP-ViT layers with similar behaviors into shallow, middle, and deep categories.<n>We revisit the visual layer selection problem in MLLMs at scale, training LLaVA-style models ranging from 1.4B to 7B parameters.<n>We find that: (1) deep layers are essential for OCR tasks; (2) shallow and middle layers substantially outperform deep layers on reasoning tasks involving counting, positioning, and object localization; and (3) a lightweight fusion of features across shallow, middle, and deep layers consistently outperforms specialized fusion baselines and single-
  arXiv  Detail & Related papers  (2025-04-30T09:07:10Z)
• Generative Multimodal Pretraining with Discrete Diffusion Timestep Tokens [66.02261367232256]
  Multimodal Large Language Models (MLLMs) aim to unify visual comprehension and generation.<n>Existing approaches rely on spatial visual tokens, where image patches are encoded and arranged according to a spatial order.<n>In this paper, we build a proper visual language by reconstructing diffusion timesteps to learn discrete visual tokens.
  arXiv  Detail & Related papers  (2025-04-20T16:14:28Z)
• LayAlign: Enhancing Multilingual Reasoning in Large Language Models via Layer-Wise Adaptive Fusion and Alignment Strategy [33.85811169010525]
  Large language models (LLMs) exhibit suboptimal performance on low-resource languages.<n>Recent approaches have leveraged multilingual encoders alongside LLMs by introducing trainable parameters connecting the two models.<n>We propose aname, a framework that integrates representations from all encoder layers.
  arXiv  Detail & Related papers  (2025-02-17T03:45:03Z)
• Toward Robust Hyper-Detailed Image Captioning: A Multiagent Approach and Dual Evaluation Metrics for Factuality and Coverage [50.84150600032693]
  Multimodal large language models (MLLMs) excel at generating highly detailed captions but often produce hallucinations.<n>We propose a multiagent approach that leverages LLM-MLLM collaboration to correct given captions.<n>Our proposed method significantly enhances the factual accuracy of captions, even improving those generated by GPT-4V.
  arXiv  Detail & Related papers  (2024-12-20T01:37:22Z)
• LLaVA-UHD v2: an MLLM Integrating High-Resolution Semantic Pyramid via Hierarchical Window Transformer [110.39467860530819]
  Vision transformers (ViTs) are widely employed in multimodal large language models (MLLMs) for visual encoding.<n>We present LLaVA-UHD v2, an MLLM with advanced perception abilities by introducing a well-designed vision-language projector.<n>Hiwin transformer enhances MLLM's ability to capture diverse multi-modal visual granularities, by incorporating our constructed high-resolution semantic pyramid.
  arXiv  Detail & Related papers  (2024-12-18T14:07:46Z)
• Accelerating Multimodal Large Language Models via Dynamic Visual-Token Exit and the Empirical Findings [66.04061083611863]
  Excessive use of visual tokens in existing Multimoal Large Language Models (MLLMs) often exhibits obvious redundancy and brings in prohibitively expensive computation.<n>We propose a simple yet effective method to improve the efficiency of MLLMs, termed dynamic visual-token exit (DyVTE)<n>DyVTE uses lightweight hyper-networks to perceive the text token status and decide the removal of all visual tokens after a certain layer.
  arXiv  Detail & Related papers  (2024-11-29T11:24:23Z)
• EMMA: Empowering Multi-modal Mamba with Structural and Hierarchical Alignment [39.870809905905325]
  We propose Empowering Multi-modal Mamba with Structural and Hierarchical Alignment (EMMA) to extract fine-grained visual information. Our model shows lower latency than other Mamba-based MLLMs and is nearly four times faster than transformer-based MLLMs of similar scale during inference.
  arXiv  Detail & Related papers  (2024-10-08T11:41:55Z)
• Boosting Multimodal Large Language Models with Visual Tokens Withdrawal for Rapid Inference [59.91176945361035]
  We introduce Visual Tokens Withdrawal (VTW), a plug-and-play module to boost MLLMs for rapid inference.<n>VTW strategically withdraws vision tokens at a certain layer, enabling only text tokens to engage in subsequent layers.<n>Our approach can cut computational overhead by over 40% across diverse multimodal tasks while maintaining performance.
  arXiv  Detail & Related papers  (2024-05-09T14:38:53Z)
• LION : Empowering Multimodal Large Language Model with Dual-Level Visual Knowledge [58.82222646803248]
  Multimodal Large Language Models (MLLMs) have endowed LLMs with the ability to perceive and understand multi-modal signals. Most of the existing MLLMs mainly adopt vision encoders pretrained on coarsely aligned image-text pairs, leading to insufficient extraction and reasoning of visual knowledge. We propose a dual-Level vIsual knedgeOwl eNhanced Multimodal Large Language Model (LION), which empowers the MLLM by injecting visual knowledge in two levels.
  arXiv  Detail & Related papers  (2023-11-20T15:56:44Z)

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.