Andrew S. Gordon

Melissa Roemmele, Andrew S. Gordon

LLMs can now perform a variety of complex writing tasks. They also excel in answering questions pertaining to natural language inference and commonsense reasoning. Composing these questions is itself a skilled writing task, so in this paper we consider LLMs as authors of commonsense assessment items. We prompt LLMs to generate items in the style of a prominent benchmark for commonsense reasoning, the Choice of Plausible Alternatives (COPA). We examine the outcome according to analyses facilitated by the LLMs and human annotation. We find that LLMs that succeed in answering the original COPA benchmark are also more successful in authoring their own items.

George Wang, Garrett Baker, Andrew Gordon et al.

Understanding how language models develop their internal computational structure is a central problem in the science of deep learning. While susceptibilities, drawn from statistical physics, offer a promising analytical tool, their full potential for visualizing network organization remains untapped. In this work, we introduce an embryological approach, applying UMAP to the susceptibility matrix to visualize the model's structural development over training. Our visualizations reveal the emergence of a clear ``body plan,'' charting the formation of known features like the induction circuit and discovering previously unknown structures, such as a ``spacing fin'' dedicated to counting space tokens. This work demonstrates that susceptibility analysis can move beyond validation to uncover novel mechanisms, providing a powerful, holistic lens for studying the developmental principles of complex neural networks.

Ruey-Cheng Chen, Reid Swanson, Andrew S. Gordon

Advances in topic modeling have yielded effective methods for characterizing the latent semantics of textual data. However, applying standard topic modeling approaches to sentence-level tasks introduces a number of challenges. In this paper, we adapt the approach of latent-Dirichlet allocation to include an additional layer for incorporating information about the sentence boundaries in documents. We show that the addition of this minimal information of document structure improves the perplexity results of a trained model.