Transcriptomic foundation models (TFMs) have recently emerged as powerful tools for analyzing gene expression in cells and tissues, supporting key tasks such as cell-type annotation, batch correction, and perturbation prediction. However, the diversity of model implementations and training strategies across recent TFMs, though promising, makes it challenging to isolate the contribution of individual design choices or evaluate their potential synergies. This hinders the field's ability to converge on best practices and limits the reproducibility of insights across studies. We present BMFM-RNA, an open-source, modular software package that unifies diverse TFM pretraining and fine-tuning objectives within a single framework. Leveraging this capability, we introduce a novel training objective, whole cell expression decoder (WCED), which captures global expression patterns using an autoencoder-like CLS bottleneck representation. In this paper, we describe the framework, supported input representations, and training objectives. We evaluated four model checkpoints pretrained on CELLxGENE using combinations of masked language modeling (MLM), WCED and multitask learning. Using the benchmarking capabilities of BMFM-RNA, we show that WCED-based models achieve performance that matches or exceeds state-of-the-art approaches like scGPT across more than a dozen datasets in both zero-shot and fine-tuning tasks. BMFM-RNA, available as part of the biomed-multi-omics project ( https://github.com/BiomedSciAI/biomed-multi-omic ), offers a reproducible foundation for systematic benchmarking and community-driven exploration of optimal TFM training strategies, enabling the development of more effective tools to leverage the latest advances in AI for understanding cell biology.
The pretraining-and-finetuning paradigm has driven significant advances across domains, such as natural language processing and computer vision, with representative pretraining paradigms such as masked language modeling and next-token prediction. However, in molecular representation learning, the task design remains largely limited to node-level denoising, which is effective at modeling local atomic environments, yet maybe insufficient for capturing the global molecular structure required by graph-level property prediction tasks, such as energy estimation and molecular regression. In this work, we present GeoRecon, a novel graph-level pretraining framework that shifts the focus from individual atoms to the molecule as an integrated whole. GeoRecon introduces a graph-level reconstruction task: during pretraining, the model is trained to generate an informative graph representation capable of accurately guiding reconstruction of the molecular geometry. This encourages the model to learn coherent, global structural features rather than isolated atomic details. Without relying on additional supervision or external data, GeoRecon outperforms node-centric baselines on multiple molecular benchmarks (e.g., QM9, MD17), demonstrating the benefit of incorporating graph-level reconstruction for learning more holistic and geometry-aware molecular embeddings.
Proteins are fundamental to biology, executing diverse functions through complex physicochemical interactions, and they hold transformative potential across medicine, materials science, and environmental applications. Protein Language Models (pLMs) aim to unlock insights from the vast space of unlabeled protein sequences by learning rich, semantic representations from primary sequences via masked language modeling. However, these models typically exhibit limited generative capacity. In this work, we introduce the Diffusion Sequence Model (DSM), a novel pLM trained with masked diffusion to enable both high-quality representation learning and generative protein design. DSM builds upon the ESM2 architecture by incorporating a masked forward diffusion process inspired by the LLaDA framework. After training, DSM is capable of generating diverse, biomimetic sequences that align with expected amino acid compositions, secondary structures, and predicted functions, even with 90\% token corruption. Furthermore, DSM's learned representations match or exceed those of similarly sized pLMs on downstream tasks. We also introduce DSM(ppi), a variant fine-tuned to generate protein binders by attending to target sequences. We demonstrate DSM(ppi)'s effectiveness on the challenging Bench-tested Binder Benchmark (BenchBB), where both DSM and DSM(ppi) produce candidates with superior predicted binding affinity compared to known binders. Our results establish masked diffusion as a powerful paradigm for unifying protein representation and generation in a single framework.
Heterogeneous graph neural networks (HGNNs) excel at capturing structural and semantic information in heterogeneous graphs (HGs), while struggling to generalize across domains and tasks. Recently, some researchers have turned to integrating HGNNs with large language models (LLMs) for more generalizable heterogeneous graph learning. However, these approaches typically extract structural information via HGNNs as HG tokens, and disparities in embedding spaces between HGNNs and LLMs have been shown to bias the LLM's comprehension of HGs. Moreover, as these HG tokens are often derived from node-level tasks, the model's ability to generalize across tasks remains limited. To this end, we propose a simple yet effective Masked Language Modeling-based method, called MLM4HG. MLM4HG introduces metapath-based textual sequences instead of HG tokens to extract structural and semantic information inherent in HGs, and designs customized textual templates to unify different graph tasks into a coherent cloze-style "mask" token prediction paradigm. Specifically, MLM4HG first converts HGs from various domains to texts based on metapaths, and subsequently combines them with the unified task texts to form a HG-based corpus. Moreover, the corpus is fed into a pretrained LM for fine-tuning with a constrained target vocabulary, enabling the fine-tuned LM to generalize to unseen target HGs. Extensive cross-domain and multi-task experiments on four real-world datasets demonstrate the superior generalization performance of MLM4HG over state-of-the-art methods in both few-shot and zero-shot scenarios. Our code is available at https://github.com/BUPT-GAMMA/MLM4HG.
Self-supervised objectives have driven major advances in NLP by leveraging large-scale unlabeled data, but such resources are scarce for many of the world's languages. Surprisingly, they have not been explored much for character-level tasks, where smaller amounts of data have the potential to be beneficial. We investigate the effectiveness of self-supervised auxiliary tasks for morphological inflection -- a character-level task highly relevant for language documentation -- in extremely low-resource settings, training encoder-decoder transformers for 19 languages and 13 auxiliary objectives. Autoencoding yields the best performance when unlabeled data is very limited, while character masked language modeling (CMLM) becomes more effective as data availability increases. Though objectives with stronger inductive biases influence model predictions intuitively, they rarely outperform standard CMLM. However, sampling masks based on known morpheme boundaries consistently improves performance, highlighting a promising direction for low-resource morphological modeling.
Self-Supervised Learning (SSL) has demonstrated strong performance in speech processing, particularly in automatic speech recognition. In this paper, we explore an SSL pretraining framework that leverages masked language modeling with targets derived from a speech recognition model. We also present chunkwise attention with dynamic chunk size sampling during pretraining to enable both full-context and streaming fine-tuning. Our experiments examine scaling with respect to model size and the amount of data. Using our method, we train the GigaAM family of models, including a state-of-the-art model for Russian speech recognition that outperforms Whisper-large-v3 by 50%. We have released our foundation and ASR models, along with the inference code, under the MIT license as open-source resources to the research community. Available at https://github.com/salute-developers/gigaam.
Inspired by the great success of Masked Language Modeling (MLM) in the natural language domain, the paradigm of self-supervised pre-training and fine-tuning has also achieved remarkable progress in the field of DNA sequence modeling. However, previous methods often relied on massive pre-training data or large-scale base models with huge parameters, imposing a significant computational burden. To address this, many works attempted to use more compact models to achieve similar outcomes but still fell short by a considerable margin. In this work, we propose a Hybrid Architecture Distillation (HAD) approach, leveraging both distillation and reconstruction tasks for more efficient and effective pre-training. Specifically, we employ the NTv2-500M as the teacher model and devise a grouping masking strategy to align the feature embeddings of visible tokens while concurrently reconstructing the invisible tokens during MLM pre-training. To validate the effectiveness of our proposed method, we conducted comprehensive experiments on the Nucleotide Transformer Benchmark and Genomic Benchmark. Compared to models with similar parameters, our model achieved excellent performance. More surprisingly, it even surpassed the distillation ceiling-teacher model on some sub-tasks, which is more than 500 $\times$ larger. Lastly, we utilize t-SNE for more intuitive visualization, which shows that our model can gain a sophisticated understanding of the intrinsic representation pattern in genomic sequences.
Protein language models (PLMs) have emerged as powerful tools to detect complex patterns of protein sequences. However, the capability of PLMs to fully capture information on protein sequences might be limited by focusing on single pre-training tasks. Although adding data modalities or supervised objectives can improve the performance of PLMs, pre-training often remains focused on denoising corrupted sequences. To push the boundaries of PLMs, our research investigated a multi-task pre-training strategy. We developed Ankh3, a model jointly optimized on two objectives: masked language modeling with multiple masking probabilities and protein sequence completion relying only on protein sequences as input. This multi-task pre-training demonstrated that PLMs can learn richer and more generalizable representations solely from protein sequences. The results demonstrated improved performance in downstream tasks, such as secondary structure prediction, fluorescence, GB1 fitness, and contact prediction. The integration of multiple tasks gave the model a more comprehensive understanding of protein properties, leading to more robust and accurate predictions.
Modern software systems generate extensive heterogeneous log data with dynamic formats, fragmented event sequences, and varying temporal patterns, making anomaly detection both crucial and challenging. To address these complexities, we propose ADALog, an adaptive, unsupervised anomaly detection framework designed for practical applicability across diverse real-world environments. Unlike traditional methods reliant on log parsing, strict sequence dependencies, or labeled data, ADALog operates on individual unstructured logs, extracts intra-log contextual relationships, and performs adaptive thresholding on normal data. The proposed approach utilizes a transformer-based, pretrained bidirectional encoder with a masked language modeling task, fine-tuned on normal logs to capture domain-specific syntactic and semantic patterns essential for accurate anomaly detection. Anomalies are identified via token-level reconstruction probabilities, aggregated into log-level scores, with adaptive percentile-based thresholding calibrated only on normal data. This allows the model to dynamically adapt to evolving system behaviors while avoiding rigid, heuristic-based thresholds common in traditional systems. We evaluate ADALog on benchmark datasets BGL, Thunderbird, and Spirit, showing strong generalization and competitive performance compared to state-of-the-art supervised and unsupervised methods. Additional ablation studies examine the effects of masking, fine-tuning, and token positioning on model behavior and interpretability.
Language models have emerged as powerful predictors of the viability of biological sequences. During training these models learn the rules of the grammar obeyed by sequences of amino acids or nucleotides. Once trained, these models can take a sequence as input and produce a likelihood score as an output; a higher likelihood implies adherence to the learned grammar and correlates with experimental fitness measurements. Here we show that in-context learning can distort the relationship between fitness and likelihood scores of sequences. This phenomenon most prominently manifests as anomalously high likelihood scores for sequences that contain repeated motifs. We use protein language models with different architectures trained on the masked language modeling objective for our experiments, and find transformer-based models to be particularly vulnerable to this effect. This behavior is mediated by a look-up operation where the model seeks the identity of the masked position by using the other copy of the repeated motif as a reference. This retrieval behavior can override the model's learned priors. This phenomenon persists for imperfectly repeated sequences, and extends to other kinds of biologically relevant features such as reversed complement motifs in RNA sequences that fold into hairpin structures.
Although transformers are widely used for various code-specific tasks, they have some significant limitations. In this paper, we investigate State Space Models (SSMs) as a potential alternative to transformers for code understanding tasks, such as code retrieval, classification, and clone detection. Previous research has already demonstrated that SSMs are more compute-efficient than transformers. In our work, we show that SSMs are also more sample-efficient and can effectively extrapolate to longer contexts (beyond the pretraining context) during fine-tuning. Through comprehensive experiments, we demonstrate that SSMs could serve as a viable alternative to transformers for code understanding tasks, while addressing some of the major limitations associated with transformers.
Target-specific peptides, such as conotoxins, exhibit exceptional binding affinity and selectivity toward ion channels and receptors. However, their therapeutic potential remains underutilized due to the limited diversity of natural variants and the labor-intensive nature of traditional optimization strategies. Here, we present CreoPep, a deep learning-based conditional generative framework that integrates masked language modeling with a progressive masking scheme to design high-affinity peptide mutants while uncovering novel structural motifs. CreoPep employs an integrative augmentation pipeline, combining FoldX-based energy screening with temperature-controlled multinomial sampling, to generate structurally and functionally diverse peptides that retain key pharmacological properties. We validate this approach by designing conotoxin inhibitors targeting the $\alpha$7 nicotinic acetylcholine receptor, achieving submicromolar potency in electrophysiological assays. Structural analysis reveals that CreoPep-generated variants engage in both conserved and novel binding modes, including disulfide-deficient forms, thus expanding beyond conventional design paradigms. Overall, CreoPep offers a robust and generalizable platform that bridges computational peptide design with experimental validation, accelerating the discovery of next-generation peptide therapeutics.
We investigate the use of LLM-generated data for continual pretraining of encoder models in specialized domains with limited training data, using the scientific domain of invasion biology as a case study. To this end, we leverage domain-specific ontologies by enriching them with LLM-generated data and pretraining the encoder model as an ontology-informed embedding model for concept definitions. To evaluate the effectiveness of this method, we compile a benchmark specifically designed for assessing model performance in invasion biology. After demonstrating substantial improvements over standard LLM pretraining, we investigate the feasibility of applying the proposed approach to domains without comprehensive ontologies by substituting ontological concepts with concepts automatically extracted from a small corpus of scientific abstracts and establishing relationships between concepts through distributional statistics. Our results demonstrate that this automated approach achieves comparable performance using only a small set of scientific abstracts, resulting in a fully automated pipeline for enhancing domain-specific understanding of small encoder models that is especially suited for application in low-resource settings and achieves performance comparable to masked language modeling pretraining on much larger datasets.
As the Information Retrieval (IR) field increasingly recognizes the importance of inclusivity, addressing the needs of low-resource languages remains a significant challenge. Transliteration between Urdu and its Romanized form, Roman Urdu, remains underexplored despite the widespread use of both scripts in South Asia. Prior work using RNNs on the Roman-Urdu-Parl dataset showed promising results but suffered from poor domain adaptability and limited evaluation. We propose a transformer-based approach using the m2m100 multilingual translation model, enhanced with masked language modeling (MLM) pretraining and fine-tuning on both Roman-Urdu-Parl and the domain-diverse Dakshina dataset. To address previous evaluation flaws, we introduce rigorous dataset splits and assess performance using BLEU, character-level BLEU, and CHRF. Our model achieves strong transliteration performance, with Char-BLEU scores of 96.37 for Urdu->Roman-Urdu and 97.44 for Roman-Urdu->Urdu. These results outperform both RNN baselines and GPT-4o Mini and demonstrate the effectiveness of multilingual transfer learning for low-resource transliteration tasks.
Lakota, a critically endangered language of the Sioux people in North America, faces significant challenges due to declining fluency among younger generations. This paper introduces LakotaBERT, the first large language model (LLM) tailored for Lakota, aiming to support language revitalization efforts. Our research has two primary objectives: (1) to create a comprehensive Lakota language corpus and (2) to develop a customized LLM for Lakota. We compiled a diverse corpus of 105K sentences in Lakota, English, and parallel texts from various sources, such as books and websites, emphasizing the cultural significance and historical context of the Lakota language. Utilizing the RoBERTa architecture, we pre-trained our model and conducted comparative evaluations against established models such as RoBERTa, BERT, and multilingual BERT. Initial results demonstrate a masked language modeling accuracy of 51% with a single ground truth assumption, showcasing performance comparable to that of English-based models. We also evaluated the model using additional metrics, such as precision and F1 score, to provide a comprehensive assessment of its capabilities. By integrating AI and linguistic methodologies, we aspire to enhance linguistic diversity and cultural resilience, setting a valuable precedent for leveraging technology in the revitalization of other endangered indigenous languages.
Vision-guided speech generation aims to produce authentic speech from facial appearance or lip motions without relying on auditory signals, offering significant potential for applications such as dubbing in filmmaking and assisting individuals with aphonia. Despite recent progress, existing methods struggle to achieve unified cross-modal alignment across semantics, timbre, and emotional prosody from visual cues, prompting us to propose Consistent Video-to-Speech (CV2S) as an extended task to enhance cross-modal consistency. To tackle emerging challenges, we introduce ImaginTalk, a novel cross-modal diffusion framework that generates faithful speech using only visual input, operating within a discrete space. Specifically, we propose a discrete lip aligner that predicts discrete speech tokens from lip videos to capture semantic information, while an error detector identifies misaligned tokens, which are subsequently refined through masked language modeling with BERT. To further enhance the expressiveness of the generated speech, we develop a style diffusion transformer equipped with a face-style adapter that adaptively customizes identity and prosody dynamics across both the channel and temporal dimensions while ensuring synchronization with lip-aware semantic features. Extensive experiments demonstrate that ImaginTalk can generate high-fidelity speech with more accurate semantic details and greater expressiveness in timbre and emotion compared to state-of-the-art baselines. Demos are shown at our project page: https://imagintalk.github.io.
Assembly code analysis and comprehension play critical roles in applications like reverse engineering, yet they face substantial challenges due to low information density and a lack of explicit syntactic structures. While traditional masked language modeling (MLM) approaches do not explicitly focus on natural language interaction, emerging decoder-focused large language models (LLMs) demonstrate partial success in binary analysis yet remain underexplored for holistic comprehension. We present Assembly Augmented Tuning, an end-to-end structural-semantic instruction tuning framework that synergizes encoder architecture with decoder-based LLMs through a projector module, where the assembly encoder extracts hardware-level structural features, the projector bridges representations with the semantic space, and the instruction-tuned LLM preserves natural language capabilities. Experimental results demonstrate three key advantages: (1) State-of-the-art performance in assembly comprehension with +39.7% Recall@1 and +17.8% MRR improvements over GPT-4-Turbo, (2) Consistent enhancements across base models (24.6-107.4% Recall@1 and 15.2-106.3% MRR on Qwen2.5-Coder, Deepseek-Coder and CodeLlama variants), and (3) Superior instruction-following capabilities (41.5%-118% improvements) with controlled code generation degradation (-8.9% to -35% across architectures).
Masked language modeling has become a widely adopted unsupervised technique to pre-train language models. However, the process of selecting tokens for masking is random, and the percentage of masked tokens is typically fixed for the entire training process. In this paper, we propose to adjust the masking ratio and to decide which tokens to mask based on a novel task-informed anti-curriculum learning scheme. First, we harness task-specific knowledge about useful and harmful tokens in order to determine which tokens to mask. Second, we propose a cyclic decaying masking ratio, which corresponds to an anti-curriculum schedule (from hard to easy). We exemplify our novel task-informed anti-curriculum by masking (TIACBM) approach across three diverse downstream tasks: sentiment analysis, text classification by topic, and authorship attribution. Our findings suggest that TIACBM enhances the ability of the model to focus on key task-relevant features, contributing to statistically significant performance gains across tasks. We release our code at https://github.com/JarcaAndrei/TIACBM.
Historically, LLMs have been trained using either autoregressive (AR) or masked language modeling (MLM) objectives, with AR models gaining dominance in recent years. However, AR models are inherently incapable of masked infilling, which is the ability to predict masked tokens between past and future context. In contrast, MLM models suffer from intrinsic computational inefficiencies during both training and inference that hinder their scalability. This work introduces MARIA (Masked and Autoregressive Infilling Architecture), a novel approach that leverages the strengths of both paradigms to achieve state-of-the-art masked infilling performance. MARIA combines a pre-trained MLM and AR model by training a linear decoder that takes their concatenated hidden states as input. This minimal modification enables the AR model to perform infilling while retaining its inherent advantages in terms of faster inference with KV caching. Our results demonstrate that MARIA significantly outperforms existing methods, namely discrete diffusion models, on masked infilling tasks.
Large Language Models (LLMs) are discovered to suffer from accurately retrieving key information. To address this, we propose Mask-Enhanced Autoregressive Prediction (MEAP), a simple yet effective training paradigm that seamlessly integrates Masked Language Modeling (MLM) into Next-Token Prediction (NTP) to enhance the latter's in-context retrieval capabilities. Specifically, MEAP first randomly masks a small fraction of input tokens and then directly performs the standard next-token prediction autoregressive using a decoder-only Transformer. MEAP eliminates the need for bidirectional attention or encoder-decoder architectures for MLM, incurring no additional computational overhead during pre-training or inference. Intensive experiments demonstrate that MEAP substantially outperforms NTP on key information retrieval and long-context reasoning tasks, while performing on par or better on commonsense reasoning tasks. The benefits of MEAP also extend to supervised fine-tuning, where it shows remarkable advantages in lost-in-the-middle scenarios, outperforming NTP by 11.77 percentage points. Our analysis indicates that MEAP's effectiveness arises from its ability to promote more distinguishable attention scores by concentrating on a reduced set of non-masked tokens. This mechanism improves the model's focus on task-relevant signals while mitigating the influence of peripheral context. These findings position MEAP as a promising training paradigm for large language models.