Language Model Pretraining

Follow this topic Ask about this topic

Language model pretraining builds general-purpose models by learning from large collections of text, commonly through next-token prediction or masked-token prediction. Researchers study how data curation, filtering, mixing, repetition, and multilingual or domain-specific corpora affect what models learn, alongside changes to tokenization, architecture, training objectives, and optimizers. Work also aims to make training cheaper through methods such as sparse attention, low-precision computation, parameter sharing, and more efficient data selection. A central tension is that lower training loss or greater data volume does not always bring better downstream abilities: data quality, diversity, exposure, and evaluation choices all matter, and their effects can vary with model scale and task.

Latest papers

448 papers72 papers in the last four weeks, up 57% on the four weeks before. 0.4% of all new papers.

Sep 3, 2026

From Zero to Hero: An Open LLM Ecosystem for Armenian

Pretraining data for Armenian, a morphologically rich and low-resource language, is scarce, and no open Armenian LLM has been released with the data and recipe needed to reproduce it. To address this gap, we curate and release two datasets. ArmWeb is an extensively validated corpus of 4.37M Armenian news documents. ArmSTEM is a parallel English-Armenian collection of 373K math and science problems with step-by-step solutions, translated into Armenian and verified through both answer-preserving LLM judgment and human evaluation. Continued pretraining of Gemma-4-E4B on these datasets yields arm-gemma-e4b, which outperforms every existing open Armenian model as well as its unadapted base, and is the first open Armenian LLM with complete training data and recipe. Our ablations show that news-only continued pretraining improves fluency while eroding knowledge, a pattern we also observe in existing Armenian models, and that a small share of verified translated STEM data reverses the loss. We further find that the largest public Armenian corpora overlap web-derived evaluation panels heavily, including a train/test self-overlap inside FineWeb-2. We openly release all data, models, and code.

Erik Arakelyan, Khatun Avetisyan, Meri Davtyan +5

1NVIDIA · 2COPA

Language Model Pretraining Language Modeling Low-Resource Language Processing

Sep 2, 2026

UE5M3 FP4 Block Scaling for Stable Language Model Pretraining

Stable 4-bit floating-point (FP4) pretraining is difficult because the E2M1 payload represents only a narrow range of magnitudes. NVIDIA's Transformer Engine \nv{} recipe addresses this with current-tensor scaling, a randomized Hadamard transform (RHT), and bfloat16 (BF16) final layers, adding work outside the FP4 matrix multiplications. We instead pair E2M1 payloads with unsigned E5M3 (\ue{}) block scales. Their wider range permits periodic tensor scaling, while our recipe applies selective stochastic rounding to backward gradients, omits RHT, and uses FP4 in all eligible internal linears. We pretrain a Nemotron-H 8B model for nearly 190 billion tokens. Compared with Transformer Engine \nv{}, the proposed block-16 recipe finishes with lower final-window training loss and, under their respective quantized-inference policies, lower validation loss measured as held-out negative log-likelihood. Its quantized-inference downstream point estimates are also higher on all three reported aggregates. A native \nv{} execution ablation that jointly removes RHT and the BF16 final-block exemption increases measured model-body token throughput by 21.2%. These results demonstrate end-to-end software-emulated \uefp{} pretraining with a simpler recipe and motivate native support for \ue{} block scaling.

Robert Hu, Carlo Luschi, Paul Balanca

GRAPHCORE RESEARCH

Stochastic Rounding Language Model Pretraining

Sep 1, 2026

Polish ModernBERT: The Long and Short of Polish Language Understanding

Encoder-only Transformers remain effective for discriminative and representation-learning tasks, yet Polish encoders still largely rely on BERT/RoBERTa-style architectures. We introduce Polish ModernBERT, a family of four Polish encoders available at Base and Large scales, each with 512-token and 8K context variants. We adapt the ModernBERT pretraining recipe through staged selection experiments and release a long-context benchmark covering legal topic classification, ideological decision-direction prediction, factual-consistency assessment over literary plot summaries, and human-rights violation assessment. Across 30 tasks, Polish ModernBERT achieves the best overall performance among the evaluated Polish encoders, reaching 83.99 and 85.11 for the Base-8K and Large-8K models, respectively. On long-context tasks, the 8K variants improve over matched Polish RoBERTa-8K baselines from 67.47 to 77.15 and from 75.88 to 78.49 at the Base and Large scales, respectively. The Base-8K model achieves this gain with 22% fewer parameters (149M vs.\ 190M). Efficiency measurements in representative inference setups show lower peak memory usage and latency than matched Polish RoBERTa baselines in both 512-token and 8K settings. Polish ModernBERT-8K-Base additionally achieves the best result on a Polish retrieval benchmark among the evaluated encoders below 300M parameters.

Michał Perełkiewicz, Sławomir Dadas, Rafał Poświata +1

National Information Processing Institute, Warsaw, Poland

Language Model Pretraining Long-Context Language Model Evaluation Long-Context Language Modeling Language Modeling

Sep 1, 2026

Probing Factual Knowledge Transfer with Training Data Interventions

Do multilingual language models transfer factual knowledge across languages during continued pretraining, or do they mostly recall facts learned directly from the target-language data? To answer this question more reliably, we propose an intervention-based framework: starting from an English-pretrained model, we continue pretraining on Persian data from which specific facts have been systematically removed at varying levels of granularity. We construct SIFT, a resource of 500 triples across 20 relations, stratified by the cultural origin of each fact's subject into general (globally prominent) and Persian-related entities, designed for both systematic fact removal from training data and evaluation, with natively written Persian cloze templates. Our results show that fact transfer is very limited: under the strictest removal condition, a large majority of English-acquired facts fail to transfer into Persian. We further show that sentence-level co-occurrence removal is insufficient to eliminate fact signal, and that easier (randomly selected) negative candidate sets substantially inflate apparent transfer by rewarding shallow associative heuristics, while performance on a harder candidate set that allows for less reliance on heuristics is much lower. Finally, we show that source-language entity frequency has a large influence, with Persian-related facts, which are orders of magnitude rarer in the English corpus, hardly transferring.

Romina Oji, Marc Braun, Marcel Bollmann +2

Department of Computer and Information Science Linköping University

Language Model Pretraining Multilingual Language Model Evaluation Factual Knowledge in Language Models Causal Interventions in Language Models Cross-Lingual Transfer

Sep 1, 2026

Pre-carved Niches: The Formation Dynamics of Modular Task Partitions in Early LLM Training

Large language models exhibit a modular internal organization that mirrors well-studied functional networks of the human brain, but how this organization forms during training is unknown: prior work has characterized finished models, not the formation process. We track formation step by step: we train a Pythia-410M model from scratch (two trajectories, bf16 and fp32) and run attribution patching at every step, alongside probes for gradient norms, effective updates, weight norms, and first-order loss decomposition across 14 tasks in four cognitive domains. Three findings. First, the modular map is pre-carved: before any learning, the dominant task pair already overlaps at ~3.6x the attribution substrate (a task-independent baseline), and its layer-0 concentration is an architecture-level constant on this model family. Second, the partition locks in through two sharp jumps whose amplitudes do not track the learning-rate schedule (the second reaching 20.4 sigma quiet-window / 6.2 sigma global), accompanied by gradient-level relative deprivation--winners receive 2.25->2.73x the loser's gradient supply, 9.5-11.5 standard deviations below a random control--that does not propagate to updates or weights. Third, deviation from the substrate appears only in the domain being learned, consistent with the hypothesis that modularity tracks learning. We close by separating the feature-level account we can defend from the mechanistic questions we cannot, and we pre-register the scale-threshold hypothesis behind our ongoing 2.8B experiments.

Guangqi Li, Yongxin Li

Zaozhuang University, Zaozhuang, Shandong, China

Language Model Pretraining Modular Neural Networks

Sep 1, 2026

Subword Segmental BabyLMs: Learning to Tokenise for Sample-Efficient Pretraining

In the standard LM training pipeline, subword tokenisation is applied as a preprocessing step. Subword segmental language modelling is an alternative paradigm in which tokenisation is learned during training, allowing the model to discover subword units that optimise its training objective. In this paper, we present our submission to the 2026 BabyLM Challenge, for which we develop two new subword segmental LMs: SubSegGPT and SubSegDeBERTa. SubSegGPT is a decoder-only model that learns tokenisation during autoregressive pretraining. SubSegDeBERTa is an encoder-based model that jointly learns to generate and tokenise masked words. We train both for the Strict and Strict-small tracks. Our top submission to Strict is SubSegDeBERTa, which achieves notable gains in zero-shot evaluation. Our top submission to Strict-small is SubSegGPT, which outperforms tokenisation-based baselines. Our results show that learnable subword tokenisation can improve sample-efficiency for BabyLM pretraining. We analyse the subword learning dynamics of our models and find that tokenisation gradually converges on subword units that balance morphological alignment and fine-grained segmentation.

Francois Meyer

Department of Computer Science University of Cape Town

Language Model Pretraining Autoregressive Language Modeling Efficient Language Model Training Language Modeling Subword Tokenization

Aug 31, 2026

REER-PT: Reverse-Engineered Reasoning for Perplexity-Guided Pre-training Data Augmentation

As language-model compute continues to scale, high-quality training data is becoming an increasingly important bottleneck. Conventional next-token prediction supervises what follows a context but leaves the intermediate reasoning behind that continuation implicit. We introduce REER-PT, a scalable framework that extends Reverse-Engineered Reasoning (REER) to raw pre-training data. REER-PT identifies continuations that are difficult to predict but can still be inferred from the preceding context, and inserts concise reasoning annotations that reconstruct the missing connection between context and continuation. Candidate annotations are generated and refined offline, with perplexity serving as the optimization signal. Constraints on length and target leakage filter out unhelpful or trivial annotations. This sparse transformation preserves the source text and remains compatible with standard next-token prediction, avoiding online reasoning rollouts during pre-training. We apply REER-PT to transform a source pre-training corpus into an augmented one. Across augmented-data, original-token, and selected-continuation comparisons, perplexity reductions range from 0.42 to 7.29, and only about 0.05% of annotation 13-grams appear verbatim in the source text. We then train two 680M-parameter models with the same architecture and training configuration on the source and augmented corpora, respectively. The augmented-data model gains up to 2.07 percentage points on several knowledge and reasoning benchmarks. Together, the perplexity analysis indicates improved continuation predictability, while the controlled pre-training experiments suggest that this augmentation can improve model performance without changing the standard pre-training objective.

Haoran Que, Jiajun Shi, Ting Huang +7

Peking University · ByteDance Seed · Work done at ByteDance Seed +1

Language Model Pretraining Synthetic Data Pretraining Synthetic Data Generation for Language Models Large Language Model-Based Data Augmentation

Aug 31, 2026

KItCAT: Knowledge Injection via Input Corruption for Auto-regressive Training

LLMs acquire vast amounts of knowledge during pre-training, but often lack the specialized knowledge needed to answer questions from niche sources such as manuals or technical documents unseen during pre-training. Continued pre-training (CPT) is widely used to inject such knowledge into model parameters. However, niche documents seldom repeat facts, making it difficult for CPT to robustly acquire such knowledge. Recent works address this by generating multiple paraphrases of the new knowledge, but paraphrasing is computationally expensive and typically requires powerful LLMs. In this work, we introduce KItCAT: Knowledge Injection via Corrupted Auto-regressive Training, a lightweight training strategy that reduces the need for paraphrasing in decoder-only LLMs. KItCAT augments standard next-token prediction by stochastically corrupting the input sequence. During training, a random subset of input tokens is replaced with other vocabulary tokens while the original next-token labels are kept unchanged. This simple intervention generates diverse training inputs from each sample, enabling large-scale data augmentation at negligible cost. We show that KItCAT consistently improves over CPT across multiple datasets and model families. Code is available at https://github.com/meghanadhpulivarthi/KItCAT.

Meghanadh Pulivarthi, Kushagra Bhushan, Vineet Kumar +5

IBM · Work done while at IBM. Currently at Amazon Books

Language Model Pretraining Knowledge Augmentation for Language Models Continued Pretraining

Aug 31, 2026

Context Staircase: Signature-Aligned Dynamics of Token Embeddings under Small Initialization

Token embeddings are the basic representational units that connect discrete tokens with continuous computation in language models. Although modern language models learn embeddings from random initialization through gradient-based training, the dynamical mechanism by which meaningful embedding structures emerge remains unclear. In this work, we identify that the evolving embedding structures are closely related to token-conditioned label and contextual distributions, which we formalize as probability signatures. We observe a progressive learning process, which we term Context Staircase: embeddings learn the low-order statistic signatures of the data before the high-order ones. More specifically, we observe that early in training they align with the simplest, context-free signature linking a token to its label, and as training proceeds, they progressively reflect signatures involving more and more context tokens. We then analyze the gradient flow of embeddings under small initialization to explain this phenomenon, deriving embedding evolution equations for feed-forward and self-attention architectures. We further extend these observations to real language-model training. Finally, we show that these embedding structures play an important role in both task learning and the incorporation of semantic structure into the embedding space. Overall, our results provide a dynamic explanation of how data statistics and architecture jointly shape token embeddings in language models, and reveal an implicit bias in the space of data statistics: training proceeds from simpler, low-order statistical relations toward increasingly complex, context-dependent ones.

Junjie Yao, Liangkai Hang, Zhi-Qin John Xu

School of Mathematical Sciences Shanghai Jiao Tong University Shanghai 200240, China · Institute of Natural Sciences, School of Mathematical Sciences Shanghai Jiao Tong University Shanghai 200240, China

Language Model Pretraining Word Embeddings Neural Network Training Dynamics Representation Geometry in Language Models LLM Training

Aug 31, 2026

Manacá-1B: An Open, Reproducible Brazilian-Portuguese Language Model and a Tokenizer-Aware, Paired Evaluation

Brazilian Portuguese remains under-served by open language models, and the few that exist are difficult to reproduce and are often compared without measures of uncertainty. We release Manacá-1B, an open decoder-only model of 1.72 billion parameters trained from scratch for Brazilian Portuguese with a fully containerized, reproducible pipeline. The pretraining is stable, with zero skipped or NaN steps and self-recovering loss spikes, and we release its full log and dynamics. We evaluate the model against nine open baselines on four Portuguese benchmarks under a single harness. Every comparison reports a standard error and a paired significance test, and the harness is validated against previously published numbers. On last-word prediction Manacá-1B is the strongest model below the 7B scale, exceeding both Tucano-1b1 and Tucano-2b4 on LAMBADA-PT with large paired margins; it is competitive on commonsense completion and near chance on multiple-choice reasoning, as are all small base models. Along the way we document a concrete evaluation pitfall: converting a SentencePiece tokenizer with case-folding normalization to the HuggingFace fast format silently drops the normalizer, routing every capitalized token to byte-fallback and depressing scores in a way that is invisible in aggregate metrics. The uncorrected tokenizer lowered LAMBADA-PT accuracy from 45.3 to 25.0; we quantify the effect and provide a one-line fix that reproduces the training tokenizer exactly. Code, raw training and evaluation logs, per-example prediction vectors, the model weights, and the corrected tokenizer are released so that every number in this paper can be recomputed.

Bruno Leonardo Santos Menezes, Carlos Leonardo Souza Cardoso, Fabio Andre Machado Porto

Laboratório Nacional de Computação Científica (LNCC), Petrópolis, Brazil

Language Model Pretraining Portuguese Language Models LLM Evaluation Open-Weight Language Models Language Modeling

Aug 30, 2026

Arkios: An Open Bilingual English-Nepali Language Model Trained From Scratch, with a Devanagari-Aware Tokenizer

We present Arkios, a 1.04B-parameter dense transformer pretrained from scratch on 150B tokens of bilingual English-Nepali text, using a custom single-file C/CUDA training stack and a Devanagari-aware byte-level BPE tokenizer built for this project. On ARC-Easy and ARC-Challenge, Arkios exceeds three comparably sized open models (Pythia-1.4B, TinyLlama-1.1B, OLMo-1B) despite an order of magnitude fewer training tokens, likely aided by a match between our educational-web-text pretraining data and ARC's grade-school-science format rather than a general capability advantage. We report full evaluation results under standard protocols, including a correction to an earlier partial-sample estimate, and findings specific to evaluating small models in a low-resource language: the standard multiple-choice-letter prompt format used by common evaluation harnesses places this model at chance on Nepali reading comprehension, and simultaneously at chance on English in the same format, which would lead a naive benchmark run to conclude the model has no Nepali ability when in fact it does. Concretely, both languages score at chance in the letter-choice format (0.240 Nepali, 0.236 English, against a chance baseline of 0.250), while scoring the answer text directly reveals genuine, English-favoring comprehension (0.306 Nepali, 0.387 English). We describe a manifest-conditioned tool-use contract introduced during instruction tuning, where tool calls are permitted only when a tool manifest is declared in context and suppressed otherwise, and report where that contract holds and where it does not. We release both the base and instruction-tuned model weights under Apache-2.0. The training code and a small privately-sourced portion of the Nepali pretraining corpus are not released; everything needed to reproduce the reported numbers from the released weights is included here.

Sajal Regmi, Siddhartha Pudasaini, Chetan Phakami Pun

Karela Technologies Inc.

Language Model Pretraining Multilingual Language Models Multilingual Language Model Evaluation Low-Resource Language Processing

Aug 30, 2026

Learning Representations through Token Prediction: Geometry, Approximation, and Downstream Guarantees

Token prediction is a central pre-training objective for modern language models. Despite its empirical success, why token prediction learns broadly useful representations remains incompletely understood. We develop a statistical framework connecting token prediction with representation geometry, encoder approximation, and downstream performance. Under a softmax prediction head, we show that accurate token prediction organizes token embeddings according to similarities between the distributions of contexts in which different token types appear, as measured by Hellinger distance, with explicit errors governed by prediction accuracy and token frequency. Meanwhile, the contextual representation provides a low-dimensional coordinate for the conditional distribution of the target token relative to these embeddings. We further introduce a self-consistency principle showing that repeated applications of a shared representation block can progressively refine the contextual representation without introducing additional block parameters. Among representations with the same prediction accuracy, this recurrent construction favors those that can be stably reconstructed from their contexts. Finally, we establish downstream guarantees for token generation, token community recovery, and classification by a linear probe, showing how prediction accuracy and recovered geometry translate into performance beyond the pre-training objective. Together, these results explain how the simple objective of predicting tokens can recover semantic geometry and produce broadly useful representations. A controlled simulation illustrates the theoretical mechanisms.

Shulei Wang

Department of Statistics, University of Illinois at Urbana-Champaign, 605 E. Springfield Ave., Champaign, IL 61820

Language Model Pretraining Next-Token Prediction Predictive Representation Learning Representation Geometry in Language Models Semantic Representation Learning

Aug 28, 2026

Curvature-Conditioned Multiscale Momentum with Sphere Constraints for LLM Pretraining

Pretraining accounts for a large fraction of the total computational cost in LLM training. However, noise-dominant gradients and the highly ill-conditioned loss landscape bring severe challenges. Although modern adaptive optimizers such as AdamW and Muon have achieved great success in large-scale pretraining, their reliance on gradient normalization offers limited mitigation of the ill-conditioned curvature. The progress along flat directions (eigen-directions of small eigenvalues), which dominates the final loss reduction, remains relatively slow. To enhance training dynamics along flat directions, we propose a curvature-conditioned multiscale momentum method with sphere constraints, delivering steady acceleration in LLM pretraining. This multiscale momentum, applied only along flat directions, pairs a slow-decay component for noise reduction with a fast-decay component for rapid curvature adaptation, harnessing their complementary strengths. Crucially, we employ a sphere constraint technique to prevent parameter inflation and excessively rapid effective learning rate decay that would otherwise arise from a naive combination. Extensive experiments show that the proposed method significantly accelerates Muon across diverse architectures (dense, MoE) and model sizes (0.12B--2.3B parameters). Theoretically, we verify the acceleration effect and provide insight into the design principles underlying the flat-direction multiscale momentum.

Shuchen Zhu, Yuxin Fang, Mingze Wang +1

ByteDance Seed · Peking University

Curvature-Aware Optimization Momentum Methods Accelerated Gradient Methods Language Model Pretraining

Aug 28, 2026

Deriving Scaling Laws for OpenEuroLLM Models: Learning Rate, Batch Size and Loss

We study the scaling behavior of learning rate and batch size in pretraining dense large language models on English-prevalent corpora. Beyond scaling jointly optimal learning rates and batch sizes, we investigate their marginal evolution with model capacity and data scale and develop a model that captures these relationships. As we employ a Warmup-Stable-Decay learning rate schedule, we further investigate the gains from learning rate annealing over a broad range of hyperparameters settings, models and data budgets, and whether the optimal learning rate and batch size transfer between the stable and decay phases. Finally, we characterize the dependence of loss on model capacity and dataset size, evaluating recently proposed scaling forms that explicitly model their interaction. We find these approaches particularly effective at capturing both undertraining and overtraining regimes across our experiments. This study establishes a first baseline and scaling procedure for the development of future OpenEuroLLM models. We open-source the complete collection of pretraining runs used in this study.

Niccolò Ajroldi, Diana Alexandra Onutu, Haider Al-Tahan +4

ELLIS Institute Tübingen · OpenEuroLLM · Eindhoven University of Technology +6

Language Model Pretraining Learning Rate Scheduling Language Model Scaling Laws Scaling Laws

Aug 27, 2026

Puro-2B: Poor Lab's Qwen2-1.5B Trained on RTX 5090 within $5090

Language model pretraining has become almost synonymous with prohibitive cost, placing it out of reach for much of the academic and open-source communities. Although strong open-source efforts already exist, including open-weight models and open-source training recipes, a cost-efficient, hardware-accessible, and open-source pretraining recipe has long been missing. Even at a small scale, training Llama-3.2-3B costs over $1.5M, and reproducing SmolLM3-3B needs over $700K. In this report, we present an open pretraining recipe designed to lower this barrier. Using this recipe, we train a collection of Puro-2B models from scratch on up to 1.4 trillion tokens with FP8 precision on consumer-grade RTX 5090 GPUs. The models in the collection differ in token budgets and selected recipe variants. Our best model is trained at a compute cost of less than $6.9K and approaches Qwen2.5-1.5B performance under our evaluation protocol. This cost efficiency is enabled by a combination of approaches, including hardware selection, low-precision training, hyperball optimization, curriculum model averaging, and the data recipe. Beyond the recipe itself, we provide two additional results. First, across the Puro-2B collection, we derive a Puro Cost Scaling Law that relates training cost to average model performance; the fitted law suggests that about $4.4K, less than $5,090, is sufficient to reach the performance of Qwen2-1.5B. Second, as an end-to-end case study, we examine how pretraining data curricula shape downstream performance after post-training. Such controlled studies are enabled by having access to the full pretraining pipeline rather than model weights alone. We release the full training recipe for Puro-2B, including data, code, and model weights under Apache 2.0 at https://huggingface.co/collections/thu-pacman/puro-2b.

Kairong Luo, Jiarui Cui, Yaorui Yin +8

Tsinghua University · Pengcheng Laboratory

Language Model Pretraining Language Model Scaling Laws Efficient Language Model Training Low-Precision Training

Aug 27, 2026

Squeezing More from Limited Data with Recursive Transformers

Pre-training under limited data requires a different view of scaling than web-scale language modeling. With a fixed data budget but relatively abundant compute, increasing parameter count helps only up to an optimal scale; beyond that point, models overfit and generalization worsens. We study this behavior across 10M-100M word pre-training budgets, two corpora, and multiple downstream evaluations, and find that optimal size depends strongly on both the data budget and the downstream target. We argue that standard Transformers scale down poorly to this setting, because embeddings consume a large fraction of the parameter budget and per-token computation is tied to representational capacity. To address this coupling, we study recursive Transformers, reusing a shared block across depth to scale compute, together with factorized embeddings to reduce vocabulary-map parameters. We train three recursive models and find that they outperform standard Transformers at 10M and 100M words, while remaining competitive with BabyLM Challenge 2025 winners.

Serdar Gülbahar, Lukas Edman, Alexander Fraser

Technical University of Munich

Language Model Pretraining Language Model Scaling Laws Representation Learning

Aug 27, 2026

TabuLM: Morphology-Aware Tabular Pre-training for Low-Resource Languages

We present TabuLM, the first language model pre-trained on Kinyarwanda tabular data. Kinyarwanda is a morphologically rich Bantu language spoken by over 12 million people in Rwanda, yet lacks any dedicated tabular representation learning resource. TabuLM extends KinyaBERT-large, a two-tier morphological transformer, with additive row, column, and cell-type embeddings and a learned table-structure attention bias that sharpens same-row and same-column attention. Pre-training uses two new objectives: Masked Cell Recovery (MCR), which masks entire cells and forces reconstruction from row and column context, and Column Type Prediction (CTP), which predicts column semantic types from observed cell values. We pre-train on 172 Rwandan government tables (~35,000 cells) from NISR, RAB, REB, and MoH open-data portals, and introduce TabQA-kin, the first native Kinyarwanda table question-answering benchmark comprising 526 QA pairs across 31 tables and four question types. TabuLM achieves 62.0% exact match on TabQA-kin, outperforming KinyaBERT-large by 5.7 EM points and all multilingual baselines (mBERT 49.3%, XLM-R 50.0%) by 11.7-12.7 points. Analysis shows that structural table embeddings are most decisive for comparison and lookup questions, while morphological awareness provides complementary gains. Our code, data, and pre-trained checkpoint are publicly available.

Ireddi Rakshitha, Devavarapu Yashwanth, Ntakirutimana Pierre

Barclays · Carnegie Mellon University

Language Model Pretraining Language Modeling Tabular Representation Learning Table QA Low-Resource Language Processing

Aug 27, 2026

Double Trouble: Bilingual Pretraining Leaves Language-Conditioned Effects in Shared-Language Representations

A concept can carry different associations across languages, while modern language models learn English alongside many other languages during pretraining. Yet comparisons among existing models cannot easily isolate how any one language changes the way these models represent English concepts because their training corpora, compute, architectures, and random seeds all differ. We study this question through a controlled experiment with 40 matched 310M-parameter decoder-only models that share an architecture, tokenizer, training recipe, and English data source. Each bilingual condition adds one of eight languages, while four experimental comparisons separately account for English exposure, total training, and English-document overlap. We align each model pair using 3,000 common English words, then measure where 1,000 held-out English concepts fall along 50 fixed semantic contrasts, such as red versus white. Across 32 experimental comparisons, English concept positions differ more between bilingual and English-only conditions than between English-only runs with different random seeds. These differences are larger in contextual states than in token embeddings and peak in middle layers. The language learned alongside English can therefore change how a model represents English concepts even when its English input representations are explicitly aligned.

Anjishnu Mukherjee, Ziwei Zhu, Antonios Anastasopoulos

George Mason University

Language Model Pretraining Multilingual Language Models Representation Geometry in Language Models

Aug 27, 2026

SPT: Skills as Pre-Training Data for Agentic Language Models

Agentic (tool-using) language models are mainly trained on tool-call traces and agent trajectories during post-training. These data provide direct behavioral supervision, but producing them requires task environments, execution, and verification, making broad tool and task coverage expensive. Publicly available skills offer another source of training data: they encode reusable tool semantics and workflows but are typically used only as inference-time context. We introduce Skill Pre-Training (SPT), a mid-training method that applies causal language modeling to SkillCorpus, a collection of public multi-file skill packages, optionally mixed with general data. To preserve relations among files within each package, we also introduce Reference Insert, a reference-aware assembly strategy that places supporting files near their mentions in the primary instruction. Experiments across multiple model scales and post-training recipes show that SPT consistently improves agentic performance over mid-training on general or trajectory data, while largely preserving general performance. Data mixture experiments show additional benefits from combining skill data with general annealing corpora. These results indicate that skill packages are a valuable data source for pre-training agentic language models.

Yufei Sun, Yudong Li, Yiming Cheng

Beijing University of Posts and Telecommunications · Tsinghua University

LLM Agent Training Tool-Augmented Language Model Agents LLM Agent Skill Learning Language Model Pretraining

Aug 26, 2026

Spectral Allocation: Why Muon Outperforms Adam, and How to Improve Muon

Orthogonal optimisers such as Muon can substantially accelerate large language model pretraining relative to Adam, yet the mechanism remains incompletely understood. We investigate this through an out-of-sample spectral probing analysis of Transformer loss landscapes. At checkpoints along real training trajectories, we decompose each momentum buffer into its singular directions and estimate the loss-optimal step size along each direction on held-out data. The resulting spectral profile is anisotropic yet stable across batches and training stages, and consistent across the optimisers and model scales: a volatile head operating at the Edge-of-Stability supports a much smaller step size than the tolerant bulk, which permits substantially larger steps. This profile provides a unified spectral allocation account of why Muon outperforms Adam, which outperforms SGD. It also exposes a limitation of Muon's uniform scaling: it still underutilises the bulk. Guided by this finding, we introduce Spectral-Aware Muon (SAMuon), which holds the head at the Muon scale and amplifies the bulk using a static spectral prior. We provide two variants: the complete SAMuon follows the measured profile using a low-rank randomised SVD and the simplified SAMuon-lite uses a two-level approximation via rank-one power iteration. Neither method adds persistent optimiser state or notable extra FLOPs beyond Muon at scale, and the idealised exact-whitening versions of both retain Muon's asymptotic convergence rate under standard assumptions. Across "modded-nanogpt" models from 124M to 1B parameters, both variants outperform tuned AdamW and Muon (Scion implementation) baselines in all evaluated model-scale and batch-size configurations. SAMuon requires 13.3% to 24.0% fewer training tokens to reach the same validation loss as Muon, while SAMuon-lite retains most of this gain with near-zero wall-clock overhead.

Xiaodong Wu, Wenyi Yu, Chao Zhang +1

Department of Engineering University of Cambridge · Department of Electronic Engineering Tsinghua University

Deep Learning Optimization Muon Optimizer Spectral Optimization Language Model Pretraining

Aug 26, 2026

One Form to Transfer Them All: Pretraining Multilingual Language Models Beyond Native Orthography

Multilingual language models transfer knowledge across languages through shared subword vocabulary, a mechanism that breaks down when related languages use different writing systems. Prior work addresses this via script equalization (romanization or IPA transcription), but direct comparisons are rare; the focus has been on encoder-only models, with most work adapting existing pretrained models. We systematically compare different input representations in autoregressive multilingual pretraining, comparing orthographic text, IPA, and romanization in a controlled setup across three scales (467M, 709M, and 1.03B) on eight languages in four typologically motivated pairs. Across a wide range of downstream tasks on seen and unseen languages, romanized pretraining yields the strongest cross-lingual transfer, and the advantage over text widens with scale. IPA improves over text in most settings but trails romanization. Surprisingly, finetuning a text-pretrained model on romanized data hurts performance on languages already covered by the base model, only marginally helping when the model lacks script coverage. Our results indicate that for multilingual models spanning typologically diverse scripts, to obtain maximum benefits, romanization should be treated as a core design choice applied at pretraining rather than a post hoc fix.

Muge Zhang, Aaron Jencks, Krishna Badikela +2

Department of Computer Science and Engineering, Ohio State University · Paul G. Allen School of Computer Science & Engineering, University of Washington

Language Model Pretraining Multilingual Language Models Autoregressive Language Modeling Cross-Lingual Transfer

Aug 26, 2026

MoganBert-TR: A Turkish Encoder Foundation Model Trained from Scratch with a CLM-to-MLM Curriculum

Turkish encoder models have adopted modern architectures while leaving the pretraining objective fixed at masked language modelling. This paper introduces MoganBert-TR, a 149M-parameter Turkish encoder foundation model trained from scratch on a language-specifically filtered corpus, together with an embedding model derived from it (MoganBert-Embed). MoganBert-TR is trained over 237.3B tokens with a two-stage CLM-to-MLM curriculum: causal language modelling first, masked language modelling for the remainder, with the transition made inside the stable phase of a WSD schedule. In a controlled ablation under an equal step budget, this design outperforms pure MLM by 2.7-3.7x on Turkish MS MARCO retrieval; the measured mechanism is embedding geometry, where a single direction absorbs 28.1% of the variance under pure MLM against 11.9% under the curriculum. Long-context extension and learning-rate decay are then split into two branches after a shared prefix: running the final portion of decay at 1024 context improves the TrGLUE average by 0.49 +/- 0.26 points across five paired seeds (p = 0.013) and beats a model-soup alternative by 0.75 points at ~4.3% additional cost. MoganBert-TR attains 78.41 on TrGLUE, the best among the Turkish ModernBERT models compared, and 77.73 on TabiBench, where it leads two of the eight categories with the largest margin on code retrieval (+3.62 points over TabiBERT). MoganBert-Embed, produced through teacher distillation and multi-signal contrastive fine-tuning, ranks first among student models on the MTEB(Turkish) overall average with 68.30 and reaches 99.5% of its 7.57B-parameter teacher's score with a 51x smaller backbone. The accompanying 50,048-token tokenizer outperforms all compared Turkish tokenizers on compression and fertility across two independent test sets. Weights, tokenizer, embedding model and evaluation code: https://huggingface.co/moganai

Furkan Yilmaz, Habibe Aleyna Tasdemir, Muhammed Faruk Gozay

Language Model Pretraining Language Modeling Masked Language Modeling Text Embedding Models Curriculum Learning

Aug 25, 2026

Effective Learning Rate Governs Loss Dynamics in Language Model Pretraining

We uncover ELR collapse in language model pretraining: learning rate (LR) and parameter norm govern loss dynamics primarily through their ratio, the effective learning rate (ELR). When ELR is matched across runs, their loss trajectories collapse throughout training despite substantially different LRs and parameter norms. Across optimizers, architectures, datasets, and model scales, mean collapse errors are typically a few x 10^-3, below the seed-to-seed variation measured in a representative configuration. Systematic ablations identify normalization design and the timescale of LR-norm variation as key determinants of collapse precision. Controlled interventions further show that weight decay and Hyperball shape loss dynamics primarily through the ELR schedules they induce. Replacing LR with ELR enables a fitted functional scaling law (FSL) to transfer across norm-control methods. The resulting ELR-based FSL also explains delayed acceleration, a recurring effect of norm control. Together, these results establish ELR as a common coordinate linking LR scheduling, norm control, and loss dynamics.

Zihan Liu, Ruiheng Zheng, Shaobo Zhang +4

Peking University · Ant Group

Neural Network Training Dynamics Language Model Pretraining Learning Rate Scheduling Language Model Scaling Laws

Aug 24, 2026

Data Mixing as Mixture Experiment: Response Surface Methodology and Optimal Design for Large Language Model Pretraining

Data mixing is a central design problem in large language model pretraining: given a fixed token budget, practitioners must decide how much data to allocate to each domain. Recent proxy-based methods address this problem by training small models on candidate mixtures, fitting a response model, and using the response to select mixtures for larger-scale training. We show that this workflow has the structure of a classical mixture experiment. Under this view, data domains are mixture components, token shares are component proportions, proxy-training runs are experimental design points, and validation loss defines a response surface over the probability simplex. We develop this formulation using sparse second-order Scheffé response-surface models and construct model-robust

\mathcal{I}

-optimal designs for proxy data-mixing experiments. Using RegMix as an empirical case study, we demonstrate how the framework can both interpret observed mixture responses and design more efficient proxy experiments. The Scheffé analysis shows that domain value is strongly relational: several domains that are weak under additive effects become favourable through pairwise interactions, especially through combinations with web-derived text. The sparse Scheffé model preserves mixture rankings across model scales and remains competitive with a flexible machine-learning predictor while providing an explicit decomposition of additive and interaction effects. In a simulation study calibrated to observed proxy-training responses, model-robust

\mathcal{I}

-optimal designs recover the relevant mixture ordering after removing about 25% of the original proxy runs. These results suggest that LLM data mixing should be treated not only as a prediction problem, but also as an experimental-design problem in which the proxy mixtures themselves can be chosen to improve statistical efficiency.

Yicheng Mao, Hongru Du

Department of Mathematics and Statistics, University of Calgary, University Drive NW, Calgary, T2N 1N4, Canada · Department of Systems and Information Engineering, University of Virginia, Charlottesville, VA, USA

Data Mixture Optimization Language Model Pretraining

Aug 24, 2026

The Geometry of Low-Resource Language Representations

The performance gap between low- and high-resource languages in LLMs is widely known, but it remains unclear which internal model factors drive these disparities. In this paper, we characterise this gap through the lens of representational geometry. Comparing the geometric properties of hidden representations across 30 languages reveals that LLM geometry is systematically related to language data availability. The most consistent effect is in final layers, where low-resource languages exhibit representational degeneration. To counter this, we investigate the effectiveness of regularisation terms to penalise degeneration during continued pretraining (CPT). Experiments monolingually adapting 9 base LLMs to 10 African languages show that geometric regularisation successfully reduces representational degeneration during CPT. For larger models, cosine similarity-based regularisation marginally improves performance over vanilla CPT, with more consistent gains on the most challenging tasks. We establish that the representational geometry of low- and high-resource languages in LLMs is measurably distinct, and that targeted geometric intervention is a viable strategy for improving CPT for low-resource languages.

Francois Meyer, Jan Buys

Department of Computer Science University of Cape Town

Low-Resource Language Processing Representation Geometry in Language Models Language Model Pretraining

Aug 24, 2026

HelaBERT: Enhancing Sinhala Language Understanding with Dual Pooling Classification Head

We present HelaBERT, a family of two BERT-based masked language models pre-trained from scratch on approximately 1 billion tokens of Sinhala text sourced from MADLAD-400, CulturaX, and a custom corpus comprising news articles, Sinhala Wikipedia, and web crawl data. HelaBERT-Small (~23.3M parameters, 6 layers) and HelaBERT-Large (~110M parameters, 12 layers) both use a SentencePiece Unigram tokenizer (vocabulary size 32,000) tailored to Sinhala's agglutinative morphology and complex script. We evaluate both models on four downstream Sinhala text classification tasks: news category classification, news source classification, sentiment analysis, and writing style classification, using 5 independent seed runs with stratified 80/20 train/test splits. We additionally propose a dual pooling classification head and evaluate it systematically across all four tasks, finding consistent improvements on sentiment analysis and a moderate gain on news category classification for HelaBERT-Small, while the standard [CLS]-linear head remains competitive on news source classification, a headline-level task with short average input length. We release both models to support further research in Sinhala NLP.

Thisen Ekanayake, Nisansa de Silva

Department of Computer Science & Engineering University of Moratuwa

Language Model Pretraining Large Language Model-Based Text Classification Language Modeling Text Classification Masked Language Modeling Low-Resource Language Processing+1

Aug 22, 2026

Learning When Not to Listen: Selective Anti-Interference Pretraining for Language Models

Language models can over-condition on irrelevant preceding text: predictions already supported by local context may still change when distant, unrelated prefix tokens are perturbed. This interference is especially consequential in long, packed, or distractor-heavy contexts, where useful evidence and irrelevant spans coexist. We propose Selective Prefix Anti-Interference Regularization (SPAR), a pretraining objective for selective anti-interference. SPAR runs the original sequence and a corrupt-prefix input in which only the far prefix is changed, then uses a short-context sufficiency gate and a gated KL objective to stabilize locally supported suffix predictions. The gate operationalizes a model-based estimate of whether the far prefix supplies additional information about the target token. Mechanism analyses show that the gate identifies locally sufficient tokens and sharply reduces prefix sensitivity on gate-selected suffix tokens. In continued training on pretrained base models, SPAR improves RULER across Qwen2.5-0.5B, Qwen2.5-3B, Llama-3.2-1B, Llama-3.1-8B, and GPT2-XL under equal counted training compute; pretraining experiments further show gains on both RULER and NoLiMa. These results show that selective anti-interference is an effective objective-level signal for robust context use.

Jinchang Zhu, Haowei He, Yi Ding +5

The Hong Kong University of Science and Technology (Guangzhou) · Institute of Artificial Intelligence (TeleAI), China Telecom · University of Macau

Language Model Pretraining Language Model Robustness Long-Context Language Modeling

Aug 19, 2026

Learned, Then Lost: A Measured Single-Example Counterfactual in Pre-training

A single training example's contribution to a finished model is normally estimated rather than measured, because measuring it takes two expensive full pre-training runs that differ in one row of one batch. We ran that counterfactual 24 times at a small scale. We trained 32 GPT-2 models at 124M parameters from scratch on OpenWebText, over four conditions and eight seeds. At step 200 of 9,536, at peak learning rate, we replaced one row of a 256-row batch with a fixed context injection carrying a 194-token passage. The three injected conditions are: 1. fluent prose with a corpus-attested subject, 2. fluent prose with a fabricated subject matched to it within 0.14% on full-batch gradient delta, and 3. random keyboard characters. The fourth condition is an uninjected twin. The passage is learned from one exposure and then decays. Fifty steps after injection, the arm that saw a passage predicts it better than the arm that did not by 0.039 and 0.044 nats of cross-entropy on the passage, at eight of eight seeds with p < 0.0001. At the final step we do not detect that difference for either passage, at p = 0.25 and p = 0.71, against minimum detectable effects of 0.025 and 0.079 nats, nor between the two passages, at p = 0.54. Every geometric measure we report is taken after that decay. Our pre-registered contrast on interpolation loss barrier is +0.0068 with p = 0.509, against a minimum detectable effect of 0.032 barrier units. Held-out cross-entropy is -0.00044 with p = 0.310. Per-layer centered kernel alignment does not detectably separate any condition at any layer. Weight displacement reaches 44.1% of the seed-to-seed Euclidean distance and is 92% settled by the midpoint of training, while the barrier reaches 3.0% of the seed-to-seed barrier. Those two figures sit roughly 15 times apart, and that is a lower bound. The injection relocates the model within its basin without moving it out.

Zachary Speck, Asa Shepard

Arizona State University · Williams College

Language Model Pretraining Memorization in Language Models Training Data Attribution Counterfactual Learning

Aug 14, 2026

Beyond the pale: Assessing prevalence and contents of extremist speech in LLM training data

Despite a strong interest on the part of the research community in the topic of trustworthy and safe AI, the composition of the text corpora that large language models (LLMs) encounter in pre- and post-training has not yet drawn much attention. In this work, we address the question of whether LLMs are exposed to unfiltered, uncontextualised extremist speech. Using several definitions of extremist speech, stemming from official documents and research literature, and an extraction pipeline combining automated text processing with expert verification, we provide a lower bound on the prevalence of extremist documents in Dolma, an open training corpus underpinning the OLMo series of models. We show that Dolma is likely to include hundreds of thousands of documents containing extremist content and hate speech of several types, including direct calls for violence, and discuss the implications of this for data curation and model pre-training.

Dmitry Nikolaev, Ashley A. Mattheis

University of Manchester

Language Model Pretraining Training Data Curation LLM Auditing

Aug 14, 2026

Scaling Domain Data Repetition in LLM Pretraining

As large language models scale, their training-token budgets must also increase to maintain an appropriate tokens-per-parameter ratio (

\mathrm{TPP}

). However, high-quality domain data is much harder to scale than general web data. As model size and the training-token budget increase, its fraction in the training mixture tends to decrease. Repeating the available high-quality data provides an effective way to counteract this dilution, but excessive repetition may lead to overfitting. We study this trade-off under practical LLM scaling, where the training-token budget grows proportionally with model size. For a fixed domain, we first find that, surprisingly at a fixed

\mathrm{TPP}

, the optimal repetition count mildly increases with model size. Across different domains, we find that the optimal repetition count is strongly negatively correlated with the final validation loss of a domain: domains with lower loss can generally benefit from more repetitions. In contrast, the amount of unique domain data is only weakly related to the optimal repetition count. These findings suggest that repetition counts tuned on smaller proxy models with the same

\mathrm{TPP}

can provide a practical estimate for larger models.

Jingwei Li, Xinran Gu, Rui Dai +5

Tsinghua University · ByteDance Seed

LLM Training Data Repetition in Language Model Training Language Model Pretraining Language Model Scaling Laws