Children learn words in a strikingly consistent order, and much of this order can be predicted from word-level properties such as frequency and concreteness. We ask whether acquisition order information is also implicitly present in the internal representations of language models, and what linguistic factors drive this. We probe 18 public checkpoints that span three architecture families at matched size, data scales from 10 million to roughly 180 billion words, and the winning systems and official baselines of the three previous BabyLM challenges. A ridge read-out of frozen hidden states predicts Wordbank age of acquisition for 619 CDI words, with the probing layer chosen on training folds only. Models trained on child-scale data match or exceed their full-scale counterparts in every family. Masked and hybrid training objectives yield more aligned spaces than purely causal training, and challenge winners are not the most aligned systems. Geometric measures of semantic structure add predictive value beyond classic predictors only for child-scale models. AoA decodability thus provides a training-free measure of developmental alignment for BabyLM evaluation.

Building an NLP training set for understudied languages, including Indigenous and endangered languages, often faces challenges due to varying degrees of resource limitations in the speaker communities. What are some reasonable approaches for training set construction in these cases? We address this question with POS tagging as the test case. Although many might consider POS tagging “a solved problem”, it remains a crucial task for descriptive linguistics and language documentation and requires laborious manual annotation. Drawing data from 12 language families, we compare in-context learning, active learning (AL), and random sampling. Our results suggest: (1) for communities whose language data can be ethically shared with an API, using only 1,000 randomly sampled tokens as prompt examples, the proprietary GPT-4.1-mini can deliver desirable performance (F1>0.83) on par with that from a training set of thousands of tokens in AL iterations; (2) in cases where communities prefer not to share data, 4,500-5,500 tokens selected from AL can yield reasonable results at a pace statistically significantly faster than random sampling, evidenced by growth curve modeling.

Readability assessment aims to automatically classify text by the level appropriate for learning readers. Traditional approaches to this task utilize a variety of linguistically motivated features paired with simple machine learning models. More recent methods have improved performance by discarding these features and utilizing deep learning models. However, it is unknown whether augmenting deep learning models with linguistically motivated features would improve performance further. This paper combines these two approaches with the goal of improving overall model performance and addressing this question. Evaluating on two large readability corpora, we find that, given sufficient training data, augmenting deep learning models with linguistically motivated features does not improve state-of-the-art performance. Our results provide preliminary evidence for the hypothesis that the state-of-the-art deep learning models represent linguistic features of the text related to readability. Future research on the nature of representations formed in these models can shed light on the learned features and their relations to linguistically motivated ones hypothesized in traditional approaches.