Mind the Heads: Topological Representation Alignment for Multimodal LLMs

Researchers propose Head-Wise Representation Alignment (HeRA), a method that aligns individual attention heads in multimodal large language models to improve cross-modal representation. HeRA uses a contrastive objective based on the Mutual K-Nearest Neighbor metric and aligns the least aligned heads, yielding performance gains across 18 benchmarks and reducing visual hallucinations.

arXiv:2606.23885v1 Announce Type: new Abstract: Representation alignment has emerged as an effective approach to improve Multimodal Large Language Models MLLMs by regularizing their internal representations toward those of an external vision encoder. However, existing methods typically align a fixed layer of the language backbone, overlooking the fine-grained structure of Transformer models. In this work, we propose Head-Wise Representation Alignment HeRA , a method that enforces cross-modal alignment at the level of individual attention heads. Our approach is grounded in the Platonic Representation Hypothesis, focusing on preserving the topological structure of representations i.e., their local neighborhood relationships across modalities. Following the Mutual K-Nearest Neighbor MKNN alignment metric, we introduce a contrastive objective that acts as a differentiable proxy for matching local structures. HeRA applies this objective during multimodal training to specific attention heads in the LLM, selected by their alignment score according to the MKNN metric. Counterintuitively, we find that aligning the least aligned heads yields the largest gains. Extensive evaluations across multiple MLLMs and 18 benchmarks demonstrate that HeRA consistently improves performance on challenging vision-centric tasks and serves as an effective regularizer against visual hallucinations by naturally curbing the over-reliance on linguistic priors. Our code is publicly released.