We present ImmuVis, a family of efficient foundation models for imaging mass cytometry (IMC), a high-throughput multiplex imaging technology that handles molecular marker measurements as image channels and enables large-scale spatial tissue profiling. Unlike natural images, multiplex imaging lacks a fixed channel space, as real-world marker sets vary across studies, violating a core assumption of standard vision backbones. To address this, ImmuVis introduces marker-adaptive hyperconvolutions that generate convolutional kernels from learned marker embeddings, enabling a single model to operate on arbitrary measured marker subsets without retraining. We pretrain ImmuVis on the largest dataset to date, IMC17M (28 cohorts, 24,405 images, 265 markers, over 17M patches), using self-supervised masked reconstruction. ImmuVis outperforms state-of-the-art baselines and ablations in virtual staining and downstream classification tasks at substantially lower compute cost than transformer-based alternatives, and is the sole model that provides calibrated uncertainty via a heteroscedastic likelihood objective. These results position ImmuVis as a practical framework for real-world IMC modeling.

Large language Models (LLMs) have achieved significant breakthroughs across diverse domains; however, they can still produce unreliable or misleading outputs. For responsible LLM application, Uncertainty Quantification (UQ) techniques are used to estimate a model's uncertainty about its outputs, indicating the likelihood that those outputs may be problematic. For LLM reasoning tasks, it is essential to estimate the uncertainty not only for the final answer, but also for the intermediate steps of the reasoning, as this can enable more fine-grained and targeted interventions. In this study, we explore what UQ metrics better reflect the LLM's "intermediate uncertainty" during reasoning. Our study reveals that an LLM's incorrect reasoning steps tend to contain tokens which are highly sensitive to the perturbations on the preceding token embeddings, indicating the model's uncertainty among multiple competing continuations. In this way, uncertain (possibly incorrect) intermediate steps can be readily identified using this sensitivity score as guidance in practice. In our experiments, we show such perturbation-based metrics achieve stronger uncertainty quantification performance compared with baselines including probability-based, sampling-based and Bayesian-based methods. Meanwhile, such metrics also enjoy good simplicity and efficiency.

Generating clinical reports that summarize abnormal patterns, diagnostic findings, and clinical interpretations from long-term EEG recordings remains labor-intensive. We present CELM, the first clinical EEG-to-Language foundation model capable of summarizing long-duration, variable-length EEG recordings and performing end-to-end clinical report generation at multiple scales. CELM integrates pretrained EEG foundation models with language models to enable scalable multimodal learning. We curate a large-scale clinical EEG dataset containing 9,922 reports paired with approximately 11,000 hours of EEG recordings from 9,048 patients to train CELM, and release the benchmark with an automated report-structuring pipeline to facilitate future research. Experimental results show that CELM consistently outperforms existing methods across all evaluation settings. Importantly, we further conduct human evaluation with clinical experts, demonstrating that CELM generates reports that are more clinically coherent, diagnostically reliable, and better aligned with expert interpretation. We release our model and benchmark construction pipeline at https://github.com/Jathurshan0330/CELM.

Training data attribution (TDA) identifies which training examples most influenced a model's prediction. Influence function methods are a theoretically grounded family of TDA methods and exploit gradients. To overcome the scalability challenge arising from gradient computation, the most popular strategy is random projection (e.g., TRAK, LoGRA). However, this still faces two bottlenecks when scaling to large training sets and high-quality attribution: \emph{(i)} storing and loading projected per-example gradients for all $N$ training examples, where query latency is dominated by I/O; and \emph{(ii)} forming the $D \times D$ inverse Hessian approximation, which costs $O(D^2)$ memory. Both bottlenecks scale with the projection dimension $D$, yet increasing $D$ is necessary for attribution quality -- creating a quality--scalability tradeoff. We introduce \textbf{LoRIF} (\textbf{Lo}w-\textbf{R}ank \textbf{I}nfluence \textbf{F}unctions), which exploits low-rank structures of gradient to address both bottlenecks. First, we store rank-$c$ factors of projected per-example gradients rather than full matrices, reducing storage and query-time I/O from $O(D)$ to $O(c\sqrt{D})$ per layer per sample. Second, we use truncated SVD with the Woodbury identity to approximate the inverse Hessian term in an $r$-dimensional subspace, reducing memory from $O(D^2)$ to $O(Dr)$. On models from 0.1B to 70B parameters trained on datasets with millions of examples, LoRIF achieves up to 20$\times$ storage reduction and query-time speedup compared to LoGRA, while matching or exceeding its attribution quality. LoRIF makes gradient-based TDA practical at frontier scale.

We present a new nonlinear dimensionality reduction method, MAPLE, that enhances UMAP by improving manifold modeling. MAPLE employs a self-supervised learning approach to more efficiently encode low-dimensional manifold geometry. Central to this approach are maximum manifold capacity representations (MMCRs), which help untangle complex manifolds by compressing variances among locally similar data points while amplifying variance among dissimilar data points. This design is particularly effective for high-dimensional data with substantial intra-cluster variance and curved manifold structures, such as biological or image data. Our qualitative and quantitative evaluations demonstrate that MAPLE can produce clearer visual cluster separations and finer subcluster resolution than UMAP while maintaining tractable computational cost.

While confidence estimation is a promising direction for mitigating hallucinations in Large Language Models (LLMs), current research overwhelmingly focuses on single-turn settings. The dynamics of model confidence in multi-turn conversations, where context accumulates and ambiguity is progressively resolved, remain largely unexplored. This work presents the first systematic study of confidence estimation in multi-turn interactions, establishing a formal evaluation framework grounded in two key desiderata: per-turn calibration and monotonicity of confidence as more information becomes available. To facilitate this, we introduce novel metrics, including a length-normalized Expected Calibration Error (InfoECE), and a new "Hinter-Guesser" paradigm for generating controlled evaluation datasets. Our experiments reveal that widely-used confidence techniques struggle with calibration and monotonicity in multi-turn dialogues. In contrast, a novel logit-based probe we introduce, P(Sufficient), proves comparatively more effective, robustly tracking evidence accumulation and distinguishing it from conversational filler. Our work provides a foundational methodology for developing more reliable and trustworthy conversational agents.

High-resolution imagery is often hindered by limitations in sensor technology, atmospheric conditions, and costs. Such challenges occur in satellite remote sensing, but also with handheld cameras, such as our smartphones. Hence, super-resolution aims to enhance the image resolution algorithmically. Since single-image super-resolution requires solving an inverse problem, such methods must exploit strong priors, e.g. learned from high-resolution training data, or be constrained by auxiliary data, e.g. by a high-resolution guide from another modality. While qualitatively pleasing, such approaches often lead to "hallucinated" structures that do not match reality. In contrast, multi-image super-resolution (MISR) aims to improve the (optical) resolution by constraining the super-resolution process with multiple views taken with sub-pixel shifts. Here, we propose SuperF, a test-time optimization approach for MISR that leverages coordinate-based neural networks, also called neural fields. Their ability to represent continuous signals with an implicit neural representation (INR) makes them an ideal fit for the MISR task. The key characteristic of our approach is to share an INR for multiple shifted low-resolution frames and to jointly optimize the frame alignment with the INR. Our approach advances related INR baselines, adopted from burst fusion for layer separation, by directly parameterizing the sub-pixel alignment as optimizable affine transformation parameters and by optimizing via a super-sampled coordinate grid that corresponds to the output resolution. Our experiments yield compelling results on simulated bursts of satellite imagery and ground-level images from handheld cameras, with upsampling factors of up to 8. A key advantage of SuperF is that this approach does not rely on any high-resolution training data.

We present GRAPE (Group Representational Position Encoding), a unified framework for positional encoding based on group actions. GRAPE unifies two families of mechanisms: (i) multiplicative rotations (Multiplicative GRAPE) in $\operatorname{SO}(d)$ and (ii) additive logit biases (Additive GRAPE) arising from unipotent actions in the general linear group $\mathrm{GL}$. In Multiplicative GRAPE, a position $n \in \mathbb{Z}$ (or $t \in \mathbb{R}$) acts as $\mathbf{G}(n) = \exp(n \, ω\, \mathbf{L})$ with a rank-2 skew-symmetric generator $\mathbf{L} \in \mathbb{R}^{d \times d}$, yielding a relative, compositional, norm-preserving map with a closed-form matrix exponential. RoPE is recovered exactly when the $d/2$ planes correspond to canonical coordinate pairs with a log-uniform spectrum. Learned commuting subspaces and compact non-commuting mixtures strictly extend this geometry to capture cross-subspace feature coupling at $O(d)$ and $O(r d)$ cost per head, respectively. In Additive GRAPE, additive logits arise from rank-1 (or low-rank) unipotent actions, recovering ALiBi and the Forgetting Transformer (FoX) as exact special cases while preserving an exact relative law and streaming cacheability. Overall, GRAPE provides a principled design space for positional geometry in long-context models, subsuming RoPE and ALiBi as special cases. Project page: https://github.com/model-architectures/GRAPE.

We consider analyzing traffic accident patterns using both road network data and satellite images aligned to road graph nodes. Previous work for predicting accident occurrences relies primarily on road network structural features while overlooking physical and environmental information from the road surface and its surroundings. In this work, we construct a large multimodal dataset spanning six U.S. states, containing nine million traffic accident records from official sources, and one million high-resolution satellite images for each node of the road network. Additionally, every node is annotated with features such as the region's weather statistics and road type (e.g., residential vs. motorway), and each edge is annotated with traffic volume information (i.e., Average Annual Daily Traffic). Utilizing this dataset, we conduct a comprehensive evaluation of multimodal learning methods that integrate both visual and network embeddings. Our findings show that integrating both data modalities improves prediction accuracy, achieving an average AUROC of $90.1\%$, a $3.7\%$ gain over graph neural network models that use only graph structures. With the improved embeddings, we conduct a causal analysis using a matching estimator to identify the key factors influencing traffic accidents. We find that accident rates rise by $24\%$ under higher precipitation, by $22\%$ on higher-speed roads such as motorways, and by $29\%$ due to seasonal patterns, after adjusting for other confounding factors. Ablation studies confirm that satellite imagery features are essential for achieving accurate prediction.

Last-layer retraining (LLR) methods -- wherein the last layer of a neural network is reinitialized and retrained on a held-out set following ERM training -- have garnered interest as an efficient approach to rectify dependence on spurious correlations and improve performance on minority groups. Surprisingly, LLR has been found to improve worst-group accuracy even when the held-out set is an imbalanced subset of the training set. We initially hypothesize that this ``unreasonable effectiveness'' of LLR is explained by its ability to mitigate neural collapse through the held-out set, resulting in the implicit bias of gradient descent benefiting robustness. Our empirical investigation does not support this hypothesis. Instead, we present strong evidence for an alternative hypothesis: that the success of LLR is primarily due to better group balance in the held-out set. We conclude by showing how the recent algorithms CB-LLR and AFR perform implicit group-balancing to elicit a robustness improvement.

Autonomous exploration of unknown environments is a key capability for mobile robots, but it is largely unsolved for robots equipped with only a single monocular camera and no dense range sensors. In this paper, we present a novel approach to monocular vision-based exploration that can safely cover large-scale unstructured indoor and outdoor 3D environments by explicitly accounting for the properties of a sparse monocular SLAM frontend in both mapping and planning. The mapping module solves the problems of sparse depth data, free-space gaps, and large depth uncertainty by oversampling free space in texture-sparse areas and keeping track of obstacle position uncertainty. The planning module handles the added free-space uncertainty through rapid replanning and perception-aware heading control. We further show that frontier-based exploration is possible with sparse monocular depth data when parallax requirements and the possibility of textureless surfaces are taken into account. We evaluate our approach extensively in diverse real-world and simulated environments, including ablation studies. To the best of the authors' knowledge, the proposed method is the first to achieve 3D monocular exploration in real-world unstructured outdoor environments. We open-source our implementation to support future research.

Understanding the training dynamics of deep neural networks remains a major open problem, with physics-inspired approaches offering promising insights. Building on this perspective, we develop a thermodynamic framework to describe the stationary distributions of stochastic gradient descent (SGD) with weight decay for scale-invariant neural networks, a setting that both reflects practical architectures with normalization layers and permits theoretical analysis. We establish analogies between training hyperparameters (e.g., learning rate, weight decay) and thermodynamic variables such as temperature, pressure, and volume. Starting with a simplified isotropic noise model, we uncover a close correspondence between SGD dynamics and ideal gas behavior, validated through theory and simulation. Extending to training of neural networks, we show that key predictions of the framework, including the behavior of stationary entropy, align closely with experimental observations. This framework provides a principled foundation for interpreting training dynamics and may guide future work on hyperparameter tuning and the design of learning rate schedulers.

Effective math tutoring requires not only solving problems but also diagnosing students' difficulties and guiding them step by step. While multimodal large language models (MLLMs) show promise, existing benchmarks largely overlook these tutoring skills. We introduce MMTutorBench, the first benchmark for AI math tutoring, consisting of 685 problems built around pedagogically significant key-steps. Each problem is paired with problem-specific rubrics that enable fine-grained evaluation across six dimensions, and structured into three tasks-Insight Discovery, Operation Formulation, and Operation Execution. We evaluate 12 leading MLLMs and find clear performance gaps between proprietary and open-source systems, substantial room compared to human tutors, and consistent trends across input variants: OCR pipelines degrade tutoring quality, few-shot prompting yields limited gains, and our rubric-based LLM-as-a-Judge proves highly reliable. These results highlight both the difficulty and diagnostic value of MMTutorBench for advancing AI tutoring.

The rising incidence of natural and human-induced disasters necessitates robust visual recognition systems capable of operating under limited labeled data conditions. However, disaster-related image classification remains challenging due to data scarcity, high intra-class variability, and domain-specific complexities in remote sensing imagery. To address these challenges, we propose the Attention Bhattacharyya Distance-based Feature Aggregation Network (ABHFA-Net), a novel few-shot learning (FSL) framework that models class prototypes as probability distributions and performs classification via Bhattacharyya distance-based comparison. Our approach integrates a spatial channel attention mechanism to enhance discrimiantive feature learning in the few-shot context and introduces a Bhattacharyya-based contrastive softmax loss for improved class separability. Extensive experiments on both benchmark datasets (CIFAR-FS, FC-100, miniImageNet, tieredImageNet) and real-world disaster datasets (AIDER, CDD, MEDIC) demonstrate the effectiveness of the proposed method. In particular, ABHFA-Net achieves 80.7% and 92.3% accuracy on CIFAR-FS under 5-way 1-shot and 5-shot settings, respectively, outperforming existing state-of-the-art methods. On disaster datasets, the model consistently improves classification performance, achieving up to 68.2% (1-shot) and 78.3% (5-shot) accuracy on AIDER, highlighting its robustness in real-world scenarios. These results establish ABHFA-Net as a strong and practical solution for few-shot disaster image classification, particularly in data-scarce and time-critical remote sensing applications. The code repository for our work is available at https://github.com/GreedYLearner1146/ABHFA-Net.

Understanding how the brain encodes visual information is a central challenge in neuroscience and machine learning. A promising approach is to reconstruct visual stimuli, essentially images, from functional Magnetic Resonance Imaging (fMRI) signals. This involves two stages: transforming fMRI signals into a latent space and then using a pretrained generative model to reconstruct images. The reconstruction quality depends on how similar the latent space is to the structure of neural activity and how well the generative model produces images from that space. Yet, it remains unclear which type of latent space best supports this transformation and how it should be organized to represent visual stimuli effectively. We present two key findings. First, fMRI signals are more similar to the text space of a language model than to either a vision based space or a joint text image space. Second, text representations and the generative model should be adapted to capture the compositional nature of visual stimuli, including objects, their detailed attributes, and relationships. Building on these insights, we propose PRISM, a model that Projects fMRI sIgnals into a Structured text space as an interMediate representation for visual stimuli reconstruction. It includes an object centric diffusion module that generates images by composing individual objects to reduce object detection errors, and an attribute relationship search module that automatically identifies key attributes and relationships that best align with the neural activity. Extensive experiments on real world datasets demonstrate that our framework outperforms existing methods, achieving up to an 8% reduction in perceptual loss. These results highlight the importance of using structured text as the intermediate space to bridge fMRI signals and image reconstruction.

Local news stations are often considered to be reliable sources of non-politicized information, particularly local concerns that residents care about. Because these stations are trusted news sources, viewers are particularly susceptible to the information they report. The Sinclair Broadcast group is a broadcasting company that has acquired many local news stations in the last decade. We investigate the effects of local news stations being acquired by Sinclair: how does coverage change? We use computational methods to investigate changes in internet content put out by local news stations before and after being acquired by Sinclair and in comparison to national news outlets. We find that there is clear evidence that local news stations report more frequently on national news at the expense of local topics, and that their coverage of polarizing national topics increases.

KV cache compression promises increased throughput and efficiency with negligible loss in performance. While the gains in throughput are indisputable and recent literature has indeed shown minimal degradation on particular benchmarks, in general the consequences of compression in realistic scenarios such as multi-instruction prompting have been insufficiently studied. In this paper, we identify several pitfalls that practitioners should be aware of when deploying KV cache compressed LLMs. We evaluate five KV cache compression methods (StreamingLLM, SnapKV, TOVA, H2O, and K-Norm) on Llama3.1 8B and Qwen2.5 14B under multi-instruction prompting with IFEval. Importantly, we show that certain instructions degrade much more rapidly with compression, effectively causing them to be completely ignored by the LLM. As a practical example, we highlight system prompt leakage as a case study, empirically demonstrating the impact of compression on leakage and general instruction-following. We identify several factors that contribute to system prompt leakage: compression method, instruction order, and KV eviction bias. We then propose simple changes to KV cache eviction policies that can reduce the impact of these factors and improve the overall performance in multi-instruction tasks.

As robots become more integrated in society, their ability to coordinate with other robots and humans on multi-modal tasks (those with multiple valid solutions) is crucial. Such behaviors can be learned from expert demonstrations via imitation learning (IL), but when expert demonstrations are multi-modal, standard IL approaches usually average across modes or collapse to a single mode, preventing effective coordination. Being inspired by diffusion models' ability to capture complex multi-modal trajectory distributions in single-agent settings, we develop a diffusion-based framework for coordinated multi-modal behavior in multi-agent systems. However, existing multi-agent diffusion approaches typically require a centralized planner or explicit communication among agents. This assumption can fail in real-world scenarios where robots must operate independently or with agents like humans that they cannot directly communicate with. Therefore, we propose MIMIC-D, a joint training with decentralized execution paradigm for multi-modal multi-agent IL via diffusion. We jointly train all agents' policies with only local information to achieve implicit coordination. In simulation and hardware experiments, our method exhibits robust multi-modal coordination behavior in various tasks and environments, improving upon state-of-the-art baselines.

The computational overhead of traditional numerical solvers for partial differential equations (PDEs) remains a critical bottleneck for large-scale parametric studies and design optimization. We introduce a Minimal-Data Parametric Neural Operator Preconditioning (MD-PNOP) framework, which establishes a new strategy for accelerating parametric PDE solvers while strictly preserving physical constraints. To address the extrapolation limitation of neural operators, parameter-induced operator difference is recast as additional source terms and incorporated into an iterative solution scheme using a pretrained neural operator. This equation-recast formulation enables systematic parameter extrapolation from a single training configuration to a broad range of unseen parameter settings without retraining. The neural operator predictions are then embedded into iterative PDE solvers as improved initial guesses, thereby reducing convergence iterations without sacrificing accuracy. Unlike purely data-driven approaches, MD-PNOP guarantees that the governing equations remain fully enforced, eliminating concerns regarding loss of physics or interpretability. The framework is architecture-agnostic and is demonstrated using both DeepONet and FNO for Boltzmann transport equation solvers in neutron transport applications. Numerical results demonstrate that neural operators trained on a single set of constant parameters successfully accelerate solutions with heterogeneous, sinusoidal, and discontinuous parameter distributions. Moreover, MD-PNOP consistently achieves approximately 50% reduction in computational time while maintaining full-order fidelity for fixed-source, single-group eigenvalue, and multigroup coupled eigenvalue problems.

Objective: Develop a cost-effective, large language model (LLM)-based pipeline for automatically extracting Review of Systems (ROS) entities from clinical notes. Materials and Methods: The pipeline extracts ROS section from the clinical note using SecTag header terminology, followed by few-shot LLMs to identify ROS entities such as diseases or symptoms, their positive/negative status and associated body systems. We implemented the pipeline using 4 open-source LLM models: llama3.1:8b, gemma3:27b, mistral3.1:24b and gpt-oss:20b. Additionally, we introduced a novel attribution algorithm that aligns LLM-identified ROS entities with their source text, addressing non-exact and synonymous matches. The evaluation was conducted on 24 general medicine notes containing 340 annotated ROS entities. Results: Open-source LLMs enable a local, cost-efficient pipeline while delivering promising performance. Larger models like Gemma, Mistral, and Gpt-oss demonstrate robust performance across three entity recognition tasks of the pipeline: ROS entity extraction, negation detection and body system classification (highest F1 score = 0.952). With the attribution algorithm, all models show improvements across key performance metrics, including higher F1 score and accuracy, along with lower error rate. Notably, the smaller Llama model also achieved promising results despite using only one-third the VRAM of larger models. Discussion and Conclusion: From an application perspective, our pipeline provides a scalable, locally deployable solution to easing the ROS documentation burden. Open-source LLMs offer a practical AI option for resource-limited healthcare settings. Methodologically, our newly developed algorithm facilitates accuracy improvements for zero- and few-shot LLMs in named entity recognition.

Planning with learned dynamics models offers a promising approach toward versatile real-world manipulation, particularly in nonprehensile settings such as pushing or rolling, where accurate analytical models are difficult to obtain. However, collecting training data for learning-based methods can be costly and inefficient, as it often relies on randomly sampled interactions that are not necessarily the most informative. Furthermore, learned models tend to exhibit high uncertainty in underexplored regions of the skill space, undermining the reliability of long-horizon planning. To address these challenges, we propose ActivePusher, a novel framework that combines residual-physics modeling with uncertainty-based active learning, to focus data acquisition on the most informative skill parameters. Additionally, ActivePusher seamlessly integrates with model-based kinodynamic planners, leveraging uncertainty estimates to bias control sampling toward more reliable actions. We evaluate our approach in both simulation and real-world environments, and demonstrate that it consistently improves data efficiency and achieves higher planning success rates in comparison to baseline methods. The source code is available at https://github.com/elpis-lab/ActivePusher.

Hallucination remains a major challenge for the safe and trustworthy deployment of large language models (LLMs) in factual content generation. Prior work has explored confidence estimation as an effective approach to hallucination detection, but often relies on post-hoc self-consistency methods that require computationally expensive sampling. Verbalized confidence offers a more efficient alternative, but existing approaches are largely limited to short-form question answering (QA) tasks and do not generalize well to open-ended generation. In this paper, we propose LoVeC (Long-form Verbalized Confidence), a novel reinforcement learning based method that trains LLMs to append an on-the-fly numerical confidence score to each generated statement during long-form generation. The confidence score serves as a direct and interpretable signal of the factuality of generation. We introduce two evaluation settings, free-form tagging and iterative tagging, to assess different verbalized confidence estimation methods. Experiments on three long-form QA datasets show that our RL-trained models achieve better calibration and generalize robustly across domains. Also, our method is highly efficient, being 20 times faster than traditional self-consistency methods while achieving better calibration.

Ill-posed inverse problems are fundamental in many domains, ranging from astrophysics to medical imaging. Emerging diffusion models provide a powerful prior for solving these problems. Existing maximum-a-posteriori (MAP) or posterior sampling approaches, however, rely on different computational approximations, leading to inaccurate or suboptimal samples. To address this issue, we introduce a new approach to solving MAP problems with diffusion model priors using a dual ascent optimization framework. Our framework achieves better image quality as measured by various metrics for image restoration problems, it is more robust to high levels of measurement noise, it is faster, and it estimates solutions that represent the observations more faithfully than the state of the art.

Foundation models are reshaping EEG analysis, yet an important problem of EEG tokenization remains a challenge. This paper presents TFM-Tokenizer, a novel tokenization framework that learns a vocabulary of time-frequency motifs from single-channel EEG signals and encodes them into discrete tokens. We propose a dual-path architecture with time-frequency masking to capture robust motif representations, and it is model-agnostic, supporting both lightweight transformers and existing foundation models for downstream tasks. Our study demonstrates three key benefits: Accuracy: Experiments on four diverse EEG benchmarks demonstrate consistent performance gains across both single- and multi-dataset pretraining settings, achieving up to $11\%$ improvement in Cohen's Kappa over strong baselines. Generalization: Moreover, as a plug-and-play component, it consistently boosts the performance of diverse foundation models, including BIOT and LaBraM. Scalability: By operating at the single-channel level rather than relying on the strict 10-20 EEG system, our method has the potential to be device-agnostic. Experiments on ear-EEG sleep staging, which differs from the pretraining data in signal format, channel configuration, recording device, and task, show that our tokenizer outperforms baselines by $14\%$. A comprehensive token analysis reveals strong class-discriminative, frequency-aware, and consistent structure, enabling improved representation quality and interpretability. Code is available at https://github.com/Jathurshan0330/TFM-Tokenizer.

The performance of an LLM depends heavily on the relevance of its training data to the downstream evaluation task. However, in practice, the data involved in an unseen evaluation task is often unknown (e.g., conversations between an LLM and a user are end-to-end encrypted). Hence, it is unclear what data are relevant for fine-tuning the LLM to maximize its performance on the specific unseen evaluation task. Instead, one can only deploy the LLM on the unseen task to gather multiple rounds of feedback on how well the model performs (e.g., user ratings). This novel setting offers a refreshing perspective towards optimizing training data mixtures via feedback from an unseen evaluation task, which prior data mixing and selection works do not consider. Our paper presents DUET, a novel global-to-local algorithm that interleaves influence function as a data selection method with Bayesian optimization to optimize data mixture via feedback from a specific unseen evaluation task. By analyzing DUET's cumulative regret, we theoretically show that DUET converges to the optimal training data mixture for an unseen task even without any data knowledge of the task. Finally, our experiments across a variety of language tasks demonstrate that DUET outperforms existing data selection and mixing methods in the unseen-task setting.

The rapid advancement of large language models (LLMs) such as GPT-3, PaLM, and Llama has significantly transformed natural language processing, showcasing remarkable capabilities in understanding and generating language. However, a fundamental bottleneck persists: these models often struggle with tasks requiring complex, multi-step reasoning, particularly in mathematical problem-solving. This deficiency stems from the critical scarcity of large-scale, high-quality, domain-specific datasets necessary for cultivating sophisticated reasoning abilities. To overcome this challenge, we introduce Template-based Data Generation (TDG), a novel and scalable paradigm that harnesses frontier LLMs (GPT-4) to automatically generate parameterized meta-templates, which in turn synthesize a virtually infinite stream of high-quality problems and solutions. Using this paradigm, we create TemplateMath Part I: TemplateGSM, a foundational dataset of over 7 million synthetically generated grade school math problems. Each problem is accompanied by a programmatically verifiable solution, offering an unprecedented level of quality at scale. This resource not only resolves the data scarcity issue for supervised fine-tuning but also provides a robust mechanism for model alignment through Reinforcement Learning with Verifiable Rewards (RLVR). Our approach elevates data augmentation by leveraging GPT-4 to generate meta-templates, ensuring diverse and complex problem structures. By providing a scalable solution to the data and verification bottleneck, TDG and TemplateGSM pave the way for a new generation of LLMs with powerful, reliable reasoning skills.

Accurate uncertainty quantification in large language models (LLMs) is essential for reliable confidence estimation, yet fine-tuned LLMs often become overconfident under limited adaptation data. Existing uncertainty methods for PEFT-based LLMs are largely post hoc, estimating uncertainty after fine-tuning rather than improving how adapters specialize to task-specific input-output relationships. We propose Functional-Level Uncertainty Quantification for Calibrated Fine-Tuning (UQ4CT), which calibrates uncertainty over the functional space induced by prompt-dependent mixtures of LoRA experts. UQ4CT implements this perspective through a mixture-of-experts fine-tuning framework, where a calibration loss aligns functional-level confidence with predictive correctness during training. Across four multiple-choice benchmarks and two open-ended generative QA tasks, UQ4CT reduces Expected Calibration Error (ECE) by over $25\%$ while preserving high accuracy. Under distribution shift, UQ4CT maintains superior calibration and competitive accuracy, demonstrating improved reliability and generalization for fine-tuned LLMs.

Backdoor attacks pose a serious security threat to large language models (LLMs), which are increasingly deployed as general-purpose assistants in safety- and privacy-critical applications. Existing LLM backdoors rely primarily on content-based triggers, requiring explicit modification of the input text. In this work, we show that this assumption is unnecessary and limiting. We introduce MetaBackdoor, a new class of backdoor attacks that exploits positional information as the trigger, without modifying textual content. Our key insight is that Transformer-based LLMs necessarily encode token positions to process ordered sequences. As a result, length-correlated positional structure is reflected in the model's internal computation and can be used as an effective non-content trigger signal. We demonstrate that even a simple length-based positional trigger is sufficient to activate stealthy backdoors. Unlike prior attacks, MetaBackdoor operates on visibly and semantically clean inputs and enables qualitatively new capabilities. We show that a backdoored LLM can be induced to disclose sensitive internal information, including proprietary system prompts, once a length condition is satisfied. We further demonstrate a self-activation scenario, where normal multi-turn interaction can move the conversation context into the trigger region and induce malicious tool-call behavior without attacker-supplied trigger text. In addition, MetaBackdoor is orthogonal to content-based backdoors and can be composed with them to create more precise and harder-to-detect activation conditions. Our results expand the threat model of LLM backdoors by revealing positional encoding as a previously overlooked attack surface. This challenges defenses that focus on detecting suspicious text and highlights the need for new defense strategies that explicitly account for positional triggers in modern LLM architectures.

Feature attribution analysis is critical for interpreting machine learning models and supporting reliable data-driven decisions. However, feature attribution measures often exhibit stochastic variation: different train--test splits, random seeds, or model-fitting procedures can produce substantially different attribution values and feature rankings. This paper proposes a framework for incorporating stochastic nature of feature attribution and a robust attribution metric, RoSHAP, for stable feature ranking based on the SHAP metric. The proposed framework models the distribution of feature attribution scores and estimates it through bootstrap resampling and kernel density estimation. We show that, under mild regularity conditions, the aggregated feature attribution score is asymptotically Gaussian, which greatly reduces the computational cost of distribution estimation. The RoSHAP summarizes the distribution of SHAP into a robust feature-ranking criterion that simultaneously rewards features that are active, strong, and stable. Through simulations and real-data experiments, the proposed framework and RoSHAP outperform standard single-run attribution measures in identifying signal features. In addition, models built using RoSHAP-selected features achieve predictive performance comparable to full-feature models while using substantially fewer predictors. The proposed RoSHAP approach improves the stability and interpretability of machine learning models, enabling reliable and consistent insights for analysis.

Moving to a new culture and adapting to a new life, as an international student, can be a stressful experience. In the US, international students face unique overlapping challenges, yet the current support ecosystem, including university support systems and informal social networks, remains largely fragmented. While conversational AI has emerged as a tool used by many (e.g., generative AI chatbots like ChatGPT and Google Gemini), we do not have a clear understanding of how international students adopt and perceive these technologies as support tools. We conducted a survey study (n=60) to map the relationship between international students' challenges and AI adoption patterns, followed by an interview study with 14 participants to identify the underlying motivations and boundaries of use. Our findings show that AI is perceived as a first-aid tool for immediate challenges, however, there is an interest in transforming AI from a tool for short-term help into a long-term support companion. By identifying where and how AI can provide long-term support, and where it is insufficient, we contribute recommendations for creating AI-powered support tailored to the unique needs of international students.