For many low-resource languages, the only available language models are large multilingual models trained on many languages simultaneously. Despite state-of-the-art performance on reasoning tasks, we find that these models still struggle with basic grammatical text generation in many languages. First, large multilingual models perform worse than bigrams for many languages (e.g. 24% of languages in XGLM 4.5B; 43% in BLOOM 7.1B) using FLORES perplexity as an evaluation metric. Second, when we train small monolingual models with only 125M parameters on 1GB or less data for 350 languages, these small models outperform large multilingual models both in perplexity and on a massively multilingual grammaticality benchmark. To facilitate future work on low-resource language modeling, we release Goldfish, a suite of over 1,000 small monolingual language models trained comparably for 350 languages. These models represent the first publicly-available monolingual language models for 215 of the languages included.

Tokenization is a critical component of the language modeling pipeline, but it is poorly understood. Tokenizers can drastically impact downstream performance, but currently there is mixed evidence about which intrinsic tokenizer evaluations are able to predict which tokenizers will result in a better language model before training the model. In this paper, we analyze approximately 200 language models trained on nine different languages. We evaluate both models and their tokenizers and test whether there is a relationship between any proposed intrinsic tokenizer evaluations and model performance, as measured by a wide variety of benchmarks. Despite the quantity and variety of proposed intrinsic tokenizer metrics, none consistently predicts downstream performance across model families, tasks, and languages. This motivates the need for novel tokenizer evaluation metrics and approaches. Our analysis of a novel linguistically grounded metric introduced in this work shows that new metrics can capture aspects of tokenizer quality that existing metrics miss, though their predictive power is likely to vary across model families and languages.

We introduce BuzzASR, a collection of language-specialized fine-tuned Whisper models adapted for automatic speech recognition (ASR) in 102 languages. Large end-to-end Transformer-based ASR models such as Whisper have revolutionized ASR, but most prominent models are highly multilingual. As a result, these models often perform poorly on languages less well-represented in their training set. While it has long been known that effective language adaptation can be achieved through simple fine-tuning on monolingual data, this strategy has only been applied to a small number of languages. We massively scale up this simple approach to 102 languages covered in the FLEURS dataset, while also implementing a more complex language adaptation strategy that integrates monolingual tokenizer replacement and data augmentation using text-only fine-tuning. BuzzASR models outperform Whisper on 77 out of 102 languages, reducing character error rates (CER) by a factor of over 2.8 on average. Our models achieve state-of-the-art CER among open-source systems on 27 of 102 languages on the combined FLEURS and Common Voice test set. Our tokenizer replacement strategy yields an average 3.4x improvement in compression rate (characters per token) over Whisper’s multilingual BPE, with gains of up to 21.7x.

Weight tying, i.e. sharing parameters between input and output embedding matrices, is common practice in language model design, yet its impact on the learned embedding space remains poorly understood. In this paper, we show that tied embedding matrices align more closely with output (unembedding) matrices than with input embeddings of comparable untied models, indicating that the shared matrix is shaped primarily for output prediction rather than input representation. This unembedding bias arises because output gradients dominate early in training. Using tuned lens analysis, we show this negatively affects early-layer computations, which contribute less effectively to the residual stream. Scaling input gradients during training reduces this bias, providing causal evidence for the role of gradient imbalance. This is mechanistic evidence that weight tying optimizes the embedding matrix for output prediction, compromising its role in input representation. These results help explain why weight tying can harm performance at scale and have implications for training smaller LLMs, where the embedding matrix contributes substantially to total parameter count.

Cross-lingual alignment in multilingual language models is typically attributed to joint training: shared parameters, mixed-language batches, or explicit alignment objectives. We ask whether monolingual models trained on non-parallel data learn alignable representations without joint training. By testing on strictly monolingual language models, such as the Goldfish model families and independently developed models from different research labs, we find three results. Correlation: these models develop alignable representational geometry across layers, with alignment strengthening as data scale, model scale, or linguistic proximity increases. Construction: a single Procrustes rotation fit on parallel sentences maps hidden states between models. Causation: the same rotation transfers functional content; patching a rotated English residual into a German model on a factual cloze flips the prediction to the donor’s capital in most cases. We confirm that cross-lingual alignment can emerge from the structure of language and the information it carries rather than from joint training, and this points to practical future directions including model stitching, merging, and modular multilingual systems built from monolingual components.

Bilingual language models (LMs) offer a controlled setting for studying how training conditions shape second-language (L2) behaviour, but prior work typically varies exposure structure, scale, and architecture at once, making it difficult to attribute effects to any single factor. We introduce **Beetle**, a controlled language model pretraining framework in which tokeniser, target language, training budget, and exposure structure are each independently manipulable, enabling systematic and comparable experimentation of training conditions. Using **Beetle**, we release 285 bilingual and 45 monolingual open-source LMs with rich checkpoints across a range of exposure schedules, data scales and first languages (L1s) to study multilingual pretraining and computational modelling of bilingualism and second language learning. Evaluating models on human bilingual and second language reading-time prediction and grammaticality judgement tasks, we find that staged and temporally structured curricula consistently improve alignment with language learner reading time compared to balanced bilingual training, with the largest gains at smaller data scales and for typologically closer language pairs. The **Beetle** models are well suited tools to help move computational psycholinguistics beyond its prevailing monolingual, English-centric focus toward models of human bilingual processing, to study cross-lingual learning dynamics, while supporting community-based development of controlled model families.

Crosslingual evaluation of language models that enables fair comparisons remains a fundamental challenge in multilingual NLP. Existing studies adopt a variety of downstream tasks and intrinsic metrics with different theoretical justifications, yet there has been little empirical investigation into whether these approaches yield meaningful crosslingual conclusions. We systematically examine crosslingual evaluation approaches using controlled monolingual language models trained on parallel data with varying tokenizer vocabulary sizes and model sizes, and further validate our findings on multilingual LLMs. We further discuss challenges in achieving comparable downstream evaluation across languages. Our results show that several widely used normalized metrics introduce crosslinguistic biases rooted in tokenization, encoding, and orthographic differences. In contrast, sentence-level negative log-likelihood computed over semantically equivalent sequences provides more meaningful and consistent crosslingual comparisons.

Language learners typically exhibit first language (L1) influence in their written second language (L2) production. We investigate whether similar patterns emerge in L2 language models (L2LMs), which are typically assessed on task-based benchmarks rather than on language use. We evaluate the use of Native Language Identification (NLI) as a method for detecting whether L2LMs exhibit human-like L1 influence. Using existing learner corpora and our novel L2 English dataset, we identify the conditions that yield the highest NLI accuracy, and show that text length but not proficiency affects performance. We then apply NLI to L2LM-generated text under various instruction-tuning and prompting conditions. We find that instruction tuning on human learner essays yields high NLI accuracy (~90%) and is necessary for detectable L1 influence. Whilst NLI accuracy is similar for L2LM and human essays, human evaluation shows that LM-generated L1 influence remains distinguishable from human writing.

Language identification (LID) is a fundamental step in curating multilingual corpora. However, LID models still perform poorly for many languages, especially on the noisy and heterogeneous web data often used to train multilingual language models. In this paper, we introduce CommonLID, a community-driven, human-annotated LID benchmark for the web domain, covering 109 languages. Many of the included languages have been previously under-served, making CommonLID a key resource for developing more representative high-quality text corpora. We show CommonLID’s value by using it, alongside five other common evaluation sets, to test eight popular LID models. We analyse our results to situate our contribution and to provide an overview of the state of the art. In particular, we highlight that existing evaluations overestimate LID accuracy for many languages in the web domain. We make CommonLID and the code used to create it available under an open, permissive license.

Language models perform differently across languages. It has been previously suggested that morphological typology may explain some of this variability (Cotterell et al., 2018). We replicate previous analyses and find additional new evidence for a performance gap between agglutinative and fusional languages, where fusional languages, such as English, tend to have better language modeling performance than morphologically more complex languages like Turkish. We then propose and test three possible causes for this performance gap: morphological alignment of tokenizers, tokenization quality, and disparities in dataset sizes and measurement. To test the morphological alignment hypothesis, we present MorphScore, a tokenizer evaluation metric, and supporting datasets for 22 languages. We find some evidence that tokenization quality explains the performance gap, but none for the role of morphological alignment. Instead we find that the performance gap is most reduced when training datasets are of equivalent size across language types, but only when scaled according to the so-called “byte-premium”—the different encoding efficiencies of different languages and orthographies. These results suggest that languages of particular morphological types are not intrinsically advantaged or disadvantaged in language modeling. Differences in performance can be attributed to disparities in dataset size. These findings bear on ongoing efforts to improve performance for low-performing and under-resourced languages.

Crosslingual transfer is crucial to contemporary language models’ multilingual capabilities, but how it occurs is not well understood. Weask what happens to a monolingual language model when it begins to be trained on a second language. Specifically, we train small bilingual models for which we control the amount of data for each language and the order of language exposure. To find evidence of shared multilingual representations, we turn to structural priming, a method used to study grammatical representations in humans. We first replicate previous crosslingual structural priming results and find that after controlling for training data quantity and language exposure, there are asymmetrical effects across language pairs and directions. We argue that this asymmetry may shape hypotheses about human structural priming effects. We also find that structural priming effects are less robust for less similar language pairs, highlighting potential limitations of crosslingual transfer learning and shared representations for typologically diverse languages.

How should text dataset sizes be compared across languages? Even for content-matched (parallel) corpora, UTF-8 encoded text can require a dramatically different number of bytes for different languages. In our work, we define the byte premium between two languages as the ratio of bytes used to encode content-matched text in those languages. We compute byte premiums for 1155 languages, and we use linear regressions to estimate byte premiums for other languages. We release a tool to obtain byte premiums for any two languages, enabling comparisons of dataset sizes across languages for more equitable multilingual model development and data practices.

The relationship between language model tokenization and performance is an open area of research. Here, we investigate how different tokenization schemes impact number agreement in Spanish plurals. We find that morphologically-aligned tokenization performs similarly to other tokenization schemes, even when induced artificially for words that would not be tokenized that way during training. We then present exploratory analyses demonstrating that language model embeddings for different plural tokenizations have similar distributions along the embedding space axis that maximally distinguishes singular and plural nouns. Our results suggest that morphologically-aligned tokenization is a viable tokenization approach, and existing models already generalize some morphological patterns to new items. However, our results indicate that morphological tokenization is not strictly required for performance.

Language models can greatly benefit from efficient tokenization. However, they still mostly utilize the classical Byte-Pair Encoding (BPE) algorithm, a simple and reliable method. BPE has been shown to cause such issues as under-trained tokens and sub-optimal compression that may affect the downstream performance. We introduce PickyBPE, a modified BPE algorithm that carries out vocabulary refinement during tokenizer training by removing merges that leave intermediate “junk” tokens. Our method improves vocabulary efficiency, eliminates under-trained tokens, and does not compromise text compression. Our experiments show that this method either improves downstream performance or does not harm it.

Multilingual language models are widely used to extend NLP systems to low-resource languages. However, concrete evidence for the effects of multilinguality on language modeling performance in individual languages remains scarce. Here, we pre-train over 10,000 monolingual and multilingual language models for over 250 languages, including multiple language families that are under-studied in NLP. We assess how language modeling performance in each language varies as a function of (1) monolingual dataset size, (2) added multilingual dataset size, (3) linguistic similarity of the added languages, and (4) model size (up to 45M parameters). We find that in moderation, adding multilingual data improves low-resource language modeling performance, similar to increasing low-resource dataset sizes by up to 33%. Improvements depend on the syntactic similarity of the added multilingual data, with marginal additional effects of vocabulary overlap. However, high-resource languages consistently perform worse in multilingual pre-training scenarios. As dataset sizes increase, adding multilingual data begins to hurt performance for both low-resource and high-resource languages, likely due to limited model capacity (the “curse of multilinguality”). These results suggest that massively multilingual pre-training may not be optimal for any languages involved, but that more targeted models can significantly improve performance.

Abstract grammatical knowledge—of parts of speech and grammatical patterns—is key to the capacity for linguistic generalization in humans. But how abstract is grammatical knowledge in large language models? In the human literature, compelling evidence for grammatical abstraction comes from structural priming. A sentence that shares the same grammatical structure as a preceding sentence is processed and produced more readily. Because confounds exist when using stimuli in a single language, evidence of abstraction is even more compelling from crosslingual structural priming, where use of a syntactic structure in one language primes an analogous structure in another language. We measure crosslingual structural priming in large language models, comparing model behavior to human experimental results from eight crosslingual experiments covering six languages, and four monolingual structural priming experiments in three non-English languages. We find evidence for abstract monolingual and crosslingual grammatical representations in the models that function similarly to those found in humans. These results demonstrate that grammatical representations in multilingual language models are not only similar across languages, but they can causally influence text produced in different languages.