Order-Level Attention Similarity Across Language Models: A Latent Commonality
This work addresses the need for systematic analysis of cross-model commonalities in language models, potentially aiding in knowledge transfer, though it is incremental in building on existing attention analysis methods.
The paper tackles the problem of identifying common context aggregation patterns across language models by introducing Order-Level Attention (OLA) and finds significant similarities in OLA across models, which correlate with syntactic knowledge; it proposes a training-free adapter method that generalizes to unseen models, enhancing their performance in experiments.
In this paper, we explore an important yet previously neglected question: Do context aggregation patterns across Language Models (LMs) share commonalities? While some works have investigated context aggregation or attention weights in LMs, they typically focus on individual models or attention heads, lacking a systematic analysis across multiple LMs to explore their commonalities. In contrast, we focus on the commonalities among LMs, which can deepen our understanding of LMs and even facilitate cross-model knowledge transfer. In this work, we introduce the Order-Level Attention (OLA) derived from the order-wise decomposition of Attention Rollout and reveal that the OLA at the same order across LMs exhibits significant similarities. Furthermore, we discover an implicit mapping between OLA and syntactic knowledge. Based on these two findings, we propose the Transferable OLA Adapter (TOA), a training-free cross-LM adapter transfer method. Specifically, we treat the OLA as a unified syntactic feature representation and train an adapter that takes OLA as input. Due to the similarities in OLA across LMs, the adapter generalizes to unseen LMs without requiring any parameter updates. Extensive experiments demonstrate that TOA's cross-LM generalization effectively enhances the performance of unseen LMs. Code is available at https://github.com/jinglin-liang/OLAS.