When Seeing Overrides Knowing: Disentangling Knowledge Conflicts in Vision-Language Models (2026.acl-long)
Copied to clipboard
| Challenge: | Vision-language models combine visual and textual information to perform complex tasks. conflicts between internal knowledge and external visual input can lead to hallucinations and unreliable predictions. |
| Approach: | They propose to use a dataset to identify attention heads that deliberately contradict internal commonsense knowledge to resolve cross-modal conflicts. |
| Outcome: | The proposed model can be manipulated to find out which visual inputs are conflicting . the model can then be orientated towards internal parametric knowledge or visual information . |
Similar Papers
Insight Over Sight: Exploring the Vision-Knowledge Conflicts in Multimodal LLMs (2025.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to mitigating vision-knowledge conflict in Large Language Models (MLLMs) are not effective and can be further scaled. |
| Approach: | They propose a framework to generate inputs to simulate and evaluate vision-knowledge conflict in Multimodal Large Language Models (MLLMs) using original images and 1,122 high-quality question-answer pairs, they propose 'a diagnostic benchmark' |
| Outcome: | The proposed framework, benchmark, and analysis contribute to the understanding and mitigation of vision-knowledge conflicts in Multimodal Large Language Models (MLLMs). |
Tug-of-War between Knowledge: Exploring and Resolving Knowledge Conflicts in Retrieval-Augmented Language Models (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing knowledge conflicts in RALMs can ensnare them in a tug-of-war between knowledge and evidence, limiting their practical applicability. |
| Approach: | They propose a method called Conflict-Disentangle Contrastive Decoding (CD2) to better calibrate the model’s confidence. |
| Outcome: | The proposed method can resolve knowledge conflicts in large language models with the help of conflict-disentangle contrast decoding (CD2) . |
Mixed Signals: Decoding VLMs’ Reasoning and Underlying Bias in Vision-Language Conflict (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Vision-language models have demonstrated impressive performance by effectively integrating visual and textual information to solve complex tasks. |
| Approach: | They build upon existing benchmarks to create five datasets containing mismatched image-text pairs and examine how they reason over visual and textual data . |
| Outcome: | The proposed model reasoned over visual and textual data in real-world applications but not in the visual and visual descriptions. |
Cutting Off the Head Ends the Conflict: A Mechanism for Interpreting and Mitigating Knowledge Conflicts in Language Models (2024.findings-acl)
Copied to clipboard
Zhuoran Jin, Pengfei Cao, Hongbang Yuan, Yubo Chen, Jiexin Xu, Huaijun Li, Xiaojian Jiang, Kang Liu, Jun Zhao
| Challenge: | Existing methods to expand internal memory boundaries of language models by providing external context can often conflict, leading to knowledge conflicts. |
| Approach: | They propose a method that prunes conflicting attention heads without updating model parameters. |
| Outcome: | The proposed method can flexibly control eight LMs to use internal memory or external context without updating model parameters. |
Can VLMs Recall Factual Associations From Visual References? (2025.findings-emnlp)
Copied to clipboard
| Challenge: | a systematic deficiency in the multimodal grounding of Vision Language Models is identified . VLMs can recall factual associations when provided a textual reference to an entity . |
| Approach: | They identify a systematic deficiency in the multimodal grounding of Vision Language Models . they show that VLMs struggle to link their internal knowledge of an entity with its image representation . |
| Outcome: | The study shows that VLMs struggle to link internal knowledge with image representations . the findings provide recommendations for future research . |
Can VLMs Actually See and Read? A Survey on Modality Collapse in Vision-Language Models (2025.findings-acl)
Copied to clipboard
| Challenge: | Vision-language models integrate textual and visual information, enabling them to process visual inputs and generate predictions. |
| Approach: | They review work on modality collapse analysis to provide insights into the reason for this unintended behavior and review probing studies for fine-grained vision-language understanding. |
| Outcome: | The proposed models can achieve competitive performance in vision-language tasks despite relying heavily on textual information and ignoring visual information. |
What Do VLMs NOTICE? A Mechanistic Interpretability Pipeline for Gaussian-Noise-free Text-Image Corruption and Evaluation (2025.naacl-long)
Copied to clipboard
| Challenge: | Vision-Language Models (VLMs) have gained prominence due to their success in solving complex cross-modal tasks. |
| Approach: | They propose a Gaussian-Noise-free pipeline for mechanistic interpretability in VLMs that introduces Semantic Image Pairs corruption, the first visual counterpart to Symmetric Token Replacement for text. |
| Outcome: | The proposed pipeline identifies a set of “universal attention heads” in BLIP and LLaVA that consistently contribute across different tasks and modalities. |
BiasDora: Exploring Hidden Biased Associations in Vision-Language Models (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing studies on social biases focus on a limited set of documented associations, such as gender-profession or race-crime. |
| Approach: | They propose to examine hidden, implicit bias associations across 9 bias dimensions by probing VLMs to uncover hidden, unexamined associations. |
| Outcome: | The proposed methods reveal that biases vary in negativity, toxicity, and extremity. |
Mitigating Hallucinations in VLMs: Enhancing Visual Attention via Head-Wise Perturbation (2026.findings-acl)
Copied to clipboard
| Challenge: | Vision–Language Models (VLMs) have demonstrated strong capabilities in tasks that require joint understanding of text and images. |
| Approach: | They propose a strategy that incorporates head-wise attention perturbation via continuous multiplicative noise coupled with a visual-guided loss focused on vision-sensitive text tokens to promote a more balanced attention distribution. |
| Outcome: | The proposed approach outperforms baseline models on three benchmarks and consistently outperformed the baseline model. |
Pixels Versus Priors: Controlling Knowledge Priors in Vision-Language Models through Visual Counterfacts (2025.emnlp-main)
Copied to clipboard
Michal Golovanevsky, William Rudman, Michael A. Lepori, Amir Bar, Ritambhara Singh, Carsten Eickhoff
| Challenge: | Multimodal Large Language Models perform well on visual question answering tasks, but it remains unclear whether their reasoning relies more on memorized world knowledge or on visual information present in the input image. |
| Approach: | They propose a dataset of visual-realistic counterfactuals that put world knowledge priors into conflict with visual input. |
| Outcome: | The proposed dataset puts world knowledge priors into conflict with visual input . it shows that model predictions shift toward visual evidence in mid-to-late layers . |