Cover Image

Not All Tokens and Heads Are Equally Important: Dual-Level Attention Intervention for Hallucination Mitigation

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Tang, Lexiang, Zhuang, Xianwei, Yang, Bang, Hu, Zhiyuan, Li, Hongxiang, Ma, Lu, Ru, Jinghan, Zou, Yuexian
Format:	Preprint
Published:	2025
Subjects:	Computer Vision and Pattern Recognition
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

VASparse: Towards Efficient Visual Hallucination Mitigation via Visual-Aware Token Sparsification
by: Zhuang, Xianwei, et al.
Published: (2025)

ATRI: Mitigating Multilingual Audio Text Retrieval Inconsistencies by Reducing Data Distribution Errors
by: Yin, Yuguo, et al.
Published: (2025)

VARGPT: Unified Understanding and Generation in a Visual Autoregressive Multimodal Large Language Model
by: Zhuang, Xianwei, et al.
Published: (2025)

Thinking by Subtraction: Confidence-Driven Contrastive Decoding for LLM Reasoning
by: Tang, Lexiang, et al.
Published: (2026)

VARGPT-v1.1: Improve Visual Autoregressive Large Unified Model via Iterative Instruction Tuning and Reinforcement Learning
by: Zhuang, Xianwei, et al.
Published: (2025)

Do we really have to filter out random noise in pre-training data for language models?
by: Ru, Jinghan, et al.
Published: (2025)

Vocabulary Hijacking in LVLMs: Unveiling Critical Attention Heads by Excluding Inert Tokens to Mitigate Hallucination
by: Chen, Yangneng, et al.
Published: (2026)

DermoGPT: Open Weights and Open Data for Morphology-Grounded Dermatological Reasoning MLLMs
by: Ru, Jinghan, et al.
Published: (2026)

Causally-Grounded Dual-Path Attention Intervention for Object Hallucination Mitigation in LVLMs
by: Yu, Liu, et al.
Published: (2025)

Not All Qubits are Utilized Equally
by: Pope, Jeremie, et al.
Published: (2025)

Not All Alumni Engage Equally
Published: (2024)

Not All Data Are Unlearned Equally
by: Krishnan, Aravind, et al.
Published: (2025)

ASK: Adaptive Self-improving Knowledge Framework for Audio Text Retrieval
by: Fu, Siyuan, et al.
Published: (2025)

Mitigating Object Hallucinations via Sentence-Level Early Intervention
by: Peng, Shangpin, et al.
Published: (2025)

Not All Tokens See Equally: Perception-Grounded Policy Optimization for Large Vision-Language Models
by: Ye, Zekai, et al.
Published: (2026)

CLAIM: Mitigating Multilingual Object Hallucination in Large Vision-Language Models with Cross-Lingual Attention Intervention
by: Ye, Zekai, et al.
Published: (2025)

Towards Spoken Language Understanding via Multi-level Multi-grained Contrastive Learning
by: Cheng, Xuxin, et al.
Published: (2024)

MAP: Mitigating Hallucinations in Large Vision-Language Models with Map-Level Attention Processing
by: Li, Chenxi, et al.
Published: (2025)

Not All Tokens Matter Equally: Dynamic In-context Vector Distillation with Decisive-Token Supervision for Long-form Medical Report Generation
by: Wu, Ning, et al.
Published: (2026)

Attention with Trained Embeddings Provably Selects Important Tokens
by: Wu, Diyuan, et al.
Published: (2025)

Cracking the Code of Hallucination in LVLMs with Vision-aware Head Divergence
by: He, Jinghan, et al.
Published: (2024)

PAINT: Paying Attention to INformed Tokens to Mitigate Hallucination in Large Vision-Language Model
by: Arif, Kazi Hasan Ibn, et al.
Published: (2025)

Mitigating Object Hallucination via Concentric Causal Attention
by: Xing, Yun, et al.
Published: (2024)

Not All Explanations for Deep Learning Phenomena Are Equally Valuable
by: Jeffares, Alan, et al.
Published: (2025)

Cross-Modal Attention Calibration for LVLM Hallucination Mitigation
by: Li, Jiaming, et al.
Published: (2025)

ICT: Image-Object Cross-Level Trusted Intervention for Mitigating Object Hallucination in Large Vision-Language Models
by: Chen, Junzhe, et al.
Published: (2024)

Mitigating Entity-Level Hallucination in Large Language Models
by: Su, Weihang, et al.
Published: (2024)

MHLA: Restoring Expressivity of Linear Attention via Token-Level Multi-Head
by: Zhang, Kewei, et al.
Published: (2026)

CAI: Caption-Sensitive Attention Intervention for Mitigating Object Hallucination in Large Vision-Language Models
by: Li, Qiming, et al.
Published: (2025)

G2L: Semantically Aligned and Uniform Video Grounding via Geodesic and Game Theory
by: Li, Hongxiang, et al.
Published: (2023)

Exploiting Auxiliary Caption for Video Grounding
by: Li, Hongxiang, et al.
Published: (2023)

Steering LVLMs via Sparse Autoencoder for Hallucination Mitigation
by: Hua, Zhenglin, et al.
Published: (2025)

Improving Audio-Text Retrieval via Hierarchical Cross-Modal Interaction and Auxiliary Captions
by: Xin, Yifei, et al.
Published: (2023)

Mitigating Modality Prior-Induced Hallucinations in Multimodal Large Language Models via Deciphering Attention Causality
by: Zhou, Guanyu, et al.
Published: (2024)

Tracing and Mitigating Hallucinations in Multimodal LLMs via Dynamic Attention Localization
by: Yang, Tiancheng, et al.
Published: (2025)

BrushEdit: All-In-One Image Inpainting and Editing
by: Li, Yaowei, et al.
Published: (2024)

RecurFormer: Not All Transformer Heads Need Self-Attention
by: Yan, Ruiqing, et al.
Published: (2024)

NCT-CRC-HE: Not All Histopathological Datasets Are Equally Useful
by: Ignatov, Andrey, et al.
Published: (2024)

Not All Noises Are Created Equally:Diffusion Noise Selection and Optimization
by: Qi, Zipeng, et al.
Published: (2024)

Seeing Far and Clearly: Mitigating Hallucinations in MLLMs with Attention Causal Decoding
by: Tang, Feilong, et al.
Published: (2025)