Beyond Logit Lens: Contextual Embeddings for Robust Hallucination Detection & Grounding in VLMs
This addresses reliability issues in multimodal AI for applications requiring accurate visual understanding, though it is incremental as it refines an existing training-free technique.
The paper tackled the problem of hallucinations in Large Multimodal Models by introducing ContextualLens, a method that uses contextual token embeddings to improve detection and grounding, achieving significant gains across diverse categories like actions and OCR.
The rapid development of Large Multimodal Models (LMMs) has significantly advanced multimodal understanding by harnessing the language abilities of Large Language Models (LLMs) and integrating modality-specific encoders. However, LMMs are plagued by hallucinations that limit their reliability and adoption. While traditional methods to detect and mitigate these hallucinations often involve costly training or rely heavily on external models, recent approaches utilizing internal model features present a promising alternative. In this paper, we critically assess the limitations of the state-of-the-art training-free technique, the logit lens, in handling generalized visual hallucinations. We introduce ContextualLens, a refined method that leverages contextual token embeddings from middle layers of LMMs. This approach significantly improves hallucination detection and grounding across diverse categories, including actions and OCR, while also excelling in tasks requiring contextual understanding, such as spatial relations and attribute comparison. Our novel grounding technique yields highly precise bounding boxes, facilitating a transition from Zero-Shot Object Segmentation to Grounded Visual Question Answering. Our contributions pave the way for more reliable and interpretable multimodal models.