What Do VLMs NOTICE? A Mechanistic Interpretability Pipeline for Noise-free Text-Image Corruption and Evaluation

Vision-Language Models (VLMs) have gained community-spanning prominence due to their ability to integrate visual and textual inputs to perform complex tasks. Despite their success, the internal decision-making processes of these models remain opaque, posing challenges in high-stakes applications. To address this, we introduce NOTICE, the first Noise-free Text-Image Corruption and Evaluation pipeline for mechanistic interpretability in VLMs. NOTICE incorporates a Semantic Minimal Pairs (SMP) framework for image corruption and Symmetric Token Replacement (STR) for text. This approach enables semantically meaningful causal mediation analysis for both modalities, providing a robust method for analyzing multimodal integration within models like BLIP. Our experiments on the SVO-Probes, MIT-States, and Facial Expression Recognition datasets reveal crucial insights into VLM decision-making, identifying the significant role of middle-layer cross-attention heads. Further, we uncover a set of ``universal cross-attention heads'' that consistently contribute across tasks and modalities, each performing distinct functions such as implicit image segmentation, object inhibition, and outlier inhibition. This work paves the way for more transparent and interpretable multimodal systems.

Outlier Dimensions Encode Task-Specific Knowledge

Representations from large language models (LLMs) are known to be dominated by a small subset of dimensions with exceedingly high variance. Previous works have argued that although ablating these outlier dimensions in LLM representations hurts downstream performance, outlier dimensions are detrimental to the representational quality of embeddings. In this study, we investigate how fine-tuning impacts outlier dimensions and show that 1) outlier dimensions that occur in pre-training persist in fine-tuned models and 2) a single outlier dimension can complete downstream tasks with a minimal error rate. Our results suggest that outlier dimensions can encode crucial task-specific knowledge and that the value of a representation in a single outlier dimension drives downstream model decisions.

Stable Anisotropic Regularization

Given the success of Large Language Models (LLMs), there has been considerable interest in studying the properties of model activations. The literature overwhelmingly agrees that LLM representations are dominated by a few ``outlier dimensions'' with exceedingly high variance and magnitude. Several studies in Natural Language Processing (NLP) have sought to mitigate the impact of such outlier dimensions and force LLMs to be isotropic (i.e., have uniform variance across all dimensions in embedding space). Isotropy is thought to be a desirable property for LLMs that improves model performance and more closely aligns textual representations with human intuition. However, many of the claims regarding isotropy in NLP have been based on the average cosine similarity of embeddings, which has recently been shown to be a flawed measure of isotropy. In this paper, we propose I-STAR: IsoScore$^{\star}$-based STable Anisotropic Regularization, a novel regularization method that can be used to increase or decrease levels of isotropy in embedding space during training. I-STAR uses IsoScore$^{\star}$, the first accurate measure of isotropy that is both differentiable and stable on mini-batch computations. In contrast to several previous works, we find that \textit{decreasing} isotropy in contextualized embeddings improves performance on the majority of tasks and models considered in this paper.

Garden-Path Traversal within GPT-2

In recent years, massive language models consisting exclusively of transformer decoders, led by the GPT-x family, have become increasingly popular. While studies have examined the behavior of these models, they tend to only focus on the output of the language model, avoiding analyzing their internal states despite such analyses being popular tools used within BERTology to study transformer encoders. We present a collection of methods for analyzing GPT-2's hidden states, and use the model's navigation of garden path sentences as a case study to demonstrate the utility of studying this model's behavior beyond its output alone. To support this analysis, we introduce a novel dataset consisting of 3 different types of garden path sentences, along with scripts to manipulate them. We find that measuring Manhattan distances and cosine similarities between hidden states shows that GPT-2 navigates these sentences more intuitively than conventional methods that predict from the model's output alone.

IsoScore: Measuring the Uniformity of Vector Space Utilization

The recent success of distributed word representations has led to an increased interest in analyzing the properties of their spatial distribution. Current metrics suggest that contextualized word embedding models do not uniformly utilize all dimensions when embedding tokens in vector space. Here we argue that existing metrics are fragile and tend to obfuscate the true spatial distribution of point clouds. To ameliorate this issue, we propose IsoScore: a novel metric which quantifies the degree to which a point cloud uniformly utilizes the ambient vector space. We demonstrate that IsoScore has several desirable properties such as mean invariance and direct correspondence to the number of dimensions used, which are properties that existing scores do not possess. Furthermore, IsoScore is conceptually intuitive and computationally efficient, making it well suited for analyzing the distribution of point clouds in arbitrary vector spaces, not necessarily limited to those of word embeddings alone. Additionally, we use IsoScore to demonstrate that a number of recent conclusions in the NLP literature that have been derived using brittle metrics of spatial distribution, such as average cosine similarity, may be incomplete or altogether inaccurate.