Back to AI News
arxiv

LeCat: Measuring LLM Category Learning from Pretraining Data

LeCat is a two-part framework for analyzing how LLMs learn category knowledge from pretraining data, distinguishing memorization, generalization, and flexible use. Experiments on open-source models and a controlled 68M model show that verbatim repetition reduces flexible learning, while diverse knowledge expressions improve robustness. Contextual flexibility remains a persistent challenge.

AI Brief

Provides concrete, recent research with actionable insights into how pretraining data expression affects category learning in LLMs.

arXiv:2611.18419v1 Announce Type: new Abstract: Humans know that airplanes are fast, that this extends to red airplanes, but that it does not hold for broken airplanes. How do LLMs acquire category knowledge that humans learn and apply effortlessly? We introduce LeCat, a two-part framework to analyze LLM's Learning of Category knowledge. First, we use LeCat to evaluate models on their ability to learn category knowledge, distinguishing between three cases: memorization of in-domain category knowledge, generalization to unseen instances, and ability to flexibly use knowledge in appropriate contexts. Then, we use LeCat to trace how this knowledge can be learned in pretraining from text that may encode category knowledge in various ways, such as verbatim repetition or exceptions, which we refer to as forms of "knowledge expression". We first assess open-source pretrained models using 199 curated category-feature pairs and analyze their training data. Our findings suggest that strengthening context and category-feature co-occurrence likelihood (PMI) improve learning for larger models and verbatim rule-like statements diminish flexible category learning. To systematically isolate these effects, we then train a 68M model from scratch with precise control over the categorical information each sees in their training data, confirming that verbatim repetition reduces flexible learning but mixture of diverse knowledge expressions lead to more robust category learning. Together, our findings highlight that while contextual flexibility is persistently difficult for models, how category knowledge is expressed in pretraining greatly impacts models' acquisition of robust category knowledge.
Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!