Towards Mitigating Hallucinations in Large Vision-Language Models by Refining Textual Embeddings
This work addresses hallucinations in large vision-language models, which is a critical issue for improving reliability in multimodal AI applications, though it is incremental as it builds on existing architectures with a simple refinement.
The paper tackled the problem of hallucinations in large vision-language models by addressing an inherent bias toward the language modality, proposing a method that refines textual embeddings with average-pooled visual features, which demonstrably improves visual grounding and significantly reduces hallucinations on established benchmarks.
In this work, we identify an inherent bias in prevailing LVLM architectures toward the language modality, largely resulting from the common practice of simply appending visual embeddings to the input text sequence. To address this, we propose a simple yet effective method that refines textual embeddings by integrating average-pooled visual features. Our approach demonstrably improves visual grounding and significantly reduces hallucinations on established benchmarks. While average pooling offers a straightforward, robust, and efficient means of incorporating visual information, we believe that more sophisticated fusion methods could further enhance visual grounding and cross-modal alignment. Given that the primary focus of this work is to highlight the modality imbalance and its impact on hallucinations -- and to show that refining textual embeddings with visual information mitigates this issue -- we leave exploration of advanced fusion strategies for future work.