MAD: Modality-Adaptive Decoding for Mitigating Cross-Modal Hallucinations in Multimodal Large Language Models

• URL: http://arxiv.org/abs/2601.21181v1
• Date: Thu, 29 Jan 2026 02:30:32 GMT
• Title: MAD: Modality-Adaptive Decoding for Mitigating Cross-Modal Hallucinations in Multimodal Large Language Models
• Authors: Sangyun Chung, Se Yeon Kim, Youngchae Chee, Yong Man Ro,
• Abstract summary: Multimodal Large Language Models (MLLMs) suffer from cross-modal hallucinations, where one modality inappropriately influences generation about another, leading to fabricated output.<n>We propose Modality-Adaptive Decoding (MAD), a training-free method that adaptively weights modality-specific decoding branches based on task requirements.<n>Our approach demonstrates that explicit modality awareness through self-assessment is crucial for robust multimodal reasoning, offering a principled extension to existing contrastive decoding methods.
• Score: 45.58164536222542
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Multimodal Large Language Models (MLLMs) suffer from cross-modal hallucinations, where one modality inappropriately influences generation about another, leading to fabricated output. This exposes a more fundamental deficiency in modality-interaction control. To address this, we propose Modality-Adaptive Decoding (MAD), a training-free method that adaptively weights modality-specific decoding branches based on task requirements. MAD leverages the model's inherent ability to self-assess modality relevance by querying which modalities are needed for each task. The extracted modality probabilities are then used to adaptively weight contrastive decoding branches, enabling the model to focus on relevant information while suppressing cross-modal interference. Extensive experiments on CMM and AVHBench demonstrate that MAD significantly reduces cross-modal hallucinations across multiple audio-visual language models (7.8\% and 2.0\% improvements for VideoLLaMA2-AV, 8.7\% and 4.7\% improvements for Qwen2.5-Omni). Our approach demonstrates that explicit modality awareness through self-assessment is crucial for robust multimodal reasoning, offering a principled extension to existing contrastive decoding methods. Our code is available at \href{https://github.com/top-yun/MAD}{https://github.com/top-yun/MAD}

Related papers

• Beyond Single Models: Mitigating Multimodal Hallucinations via Adaptive Token Ensemble Decoding [41.828387997311474]
  Large Vision-Language Models (LVLMs) have recently achieved impressive results in multimodal tasks such as image captioning and visual question answering.<n>They remain prone to object hallucination -- generating descriptions of nonexistent or misidentified objects.<n>We propose Adaptive Token Ensemble Decoding (ATED), a training-free, token-level ensemble framework that mitigates hallucination by aggregating predictions from multiple LVLMs during inference.
  arXiv  Detail & Related papers  (2025-10-21T06:11:24Z)
• IMG: Calibrating Diffusion Models via Implicit Multimodal Guidance [74.89810064807142]
  Implicit Multimodal Guidance (IMG) is a novel re-generation-based multimodal alignment framework.<n>IMG identifies misalignments and formulates the re-alignment goal into a trainable objective.<n>IMG acts as a flexible plug-and-play adapter, seamlessly enhancing prior finetuning-based alignment methods.
  arXiv  Detail & Related papers  (2025-09-30T13:27:03Z)
• MoCa: Modality-aware Continual Pre-training Makes Better Bidirectional Multimodal Embeddings [75.0617088717528]
  MoCa is a framework for transforming pre-trained VLM backbones into effective bidirectional embedding models.<n>MoCa consistently improves performance across MMEB and ViDoRe-v2 benchmarks, achieving new state-of-the-art results.
  arXiv  Detail & Related papers  (2025-06-29T06:41:00Z)
• Fork-Merge Decoding: Enhancing Multimodal Understanding in Audio-Visual Large Language Models [35.86252379746625]
  The goal of this work is to enhance balanced multimodal understanding in audio-visual large language models (AV-LLMs)<n>In current AV-LLMs, audio and video features are typically processed jointly in the decoder.<n>We propose Fork-Merge Decoding (FMD), a simple yet effective inference-time strategy that requires no additional training or architectural modifications.
  arXiv  Detail & Related papers  (2025-05-27T08:22:56Z)
• ClearSight: Visual Signal Enhancement for Object Hallucination Mitigation in Multimodal Large language Models [28.24397677839652]
  Contrastive decoding strategies are widely used to mitigate object hallucinations in multimodal large language models (MLLMs)<n>We propose Visual Amplification Fusion (VAF), a plug-and-play technique that enhances attention to visual signals within the model's middle layers.<n>VAF significantly reduces hallucinations across various MLLMs without affecting inference speed, while maintaining coherence and accuracy in generated outputs.
  arXiv  Detail & Related papers  (2025-03-17T12:30:40Z)
• M$^3$amba: CLIP-driven Mamba Model for Multi-modal Remote Sensing Classification [23.322598623627222]
  M$3$amba is a novel end-to-end CLIP-driven Mamba model for multi-modal fusion.<n>We introduce CLIP-driven modality-specific adapters to achieve a comprehensive semantic understanding of different modalities.<n>Experiments have shown that M$3$amba has an average performance improvement of at least 5.98% compared with the state-of-the-art methods.
  arXiv  Detail & Related papers  (2025-03-09T05:06:47Z)
• Unsupervised Modality Adaptation with Text-to-Image Diffusion Models for Semantic Segmentation [54.96563068182733]
  We propose Modality Adaptation with text-to-image Diffusion Models (MADM) for semantic segmentation task. MADM utilizes text-to-image diffusion models pre-trained on extensive image-text pairs to enhance the model's cross-modality capabilities. We show that MADM achieves state-of-the-art adaptation performance across various modality tasks, including images to depth, infrared, and event modalities.
  arXiv  Detail & Related papers  (2024-10-29T03:49:40Z)
• LLMs Can Evolve Continually on Modality for X-Modal Reasoning [62.2874638875554]
  Existing methods rely heavily on modal-specific pretraining and joint-modal tuning, leading to significant computational burdens when expanding to new modalities. We propose PathWeave, a flexible and scalable framework with modal-Path sWitching and ExpAnsion abilities. PathWeave performs comparably to state-of-the-art MLLMs while concurrently reducing parameter training burdens by 98.73%.
  arXiv  Detail & Related papers  (2024-10-26T13:19:57Z)
• Large Language Models for Multimodal Deformable Image Registration [50.91473745610945]
  We propose a novel coarse-to-fine MDIR framework,LLM-Morph, for aligning the deep features from different modal medical images. Specifically, we first utilize a CNN encoder to extract deep visual features from cross-modal image pairs, then we use the first adapter to adjust these tokens, and use LoRA in pre-trained LLMs to fine-tune their weights. Third, for the alignment of tokens, we utilize other four adapters to transform the LLM-encoded tokens into multi-scale visual features, generating multi-scale deformation fields and facilitating the coarse-to-fine MDIR task
  arXiv  Detail & Related papers  (2024-08-20T09:58:30Z)
• What Makes for Robust Multi-Modal Models in the Face of Missing Modalities? [35.19295402483624]
  We model the scenarios of multi-modal models encountering missing modalities from an information-theoretic perspective. We introduce Uni-Modal Ensemble with Missing Modality Adaptation (UME-MMA) UME-MMA employs uni-modal pre-trained weights for the multi-modal model to enhance feature extraction and utilizes missing modality data augmentation techniques to better adapt to situations with missing modalities.
  arXiv  Detail & Related papers  (2023-10-10T07:47:57Z)
• VERITE: A Robust Benchmark for Multimodal Misinformation Detection Accounting for Unimodal Bias [17.107961913114778]
  multimodal misinformation is a growing problem on social media platforms. In this study, we investigate and identify the presence of unimodal bias in widely-used MMD benchmarks. We introduce a new method -- termed Crossmodal HArd Synthetic MisAlignment (CHASMA) -- for generating realistic synthetic training data.
  arXiv  Detail & Related papers  (2023-04-27T12:28:29Z)

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.