DOCS: Quantifying Weight Similarity for Deeper Insights into Large Language Models
This research provides tools for deeper insights into LLM architectures, potentially aiding in developing more efficient and interpretable models, though it is incremental in nature.
The authors tackled the problem of analyzing weight similarity in Large Language Models (LLMs) by introducing the Distribution of Cosine Similarity (DOCS) index, which revealed that adjacent layers often show high similarity and form clusters, indicating depth-wise functional specialization.
We introduce a novel index, the Distribution of Cosine Similarity (DOCS), for quantitatively assessing the similarity between weight matrices in Large Language Models (LLMs), aiming to facilitate the analysis of their complex architectures. Leveraging DOCS, our analysis uncovers intriguing patterns in the latest open-source LLMs: adjacent layers frequently exhibit high weight similarity and tend to form clusters, suggesting depth-wise functional specialization. Additionally, we prove that DOCS is theoretically effective in quantifying similarity for orthogonal matrices, a crucial aspect given the prevalence of orthogonal initializations in LLMs. This research contributes to a deeper understanding of LLM architecture and behavior, offering tools with potential implications for developing more efficient and interpretable models.