Numerical reasoning over documents, which demands both contextual understanding and logical inference, is challenging for low-capacity local models deployed on computation-constrained devices. Although such complex reasoning queries could be routed to powerful remote models like GPT-4, exposing local data raises significant data leakage concerns. Existing mitigation methods generate problem descriptions or examples for remote assistance. However, the inherent complexity of numerical reasoning hinders the local model from generating logically equivalent queries and accurately inferring answers with remote guidance. In this paper, we present a model collaboration framework with two key innovations: (1) a context-aware synthesis strategy that shifts the query topics while preserving reasoning patterns; and (2) a tool-based answer reconstruction approach that reuses the remote-generated plug-and-play solution with code snippets. Experimental results demonstrate that our method achieves better reasoning accuracy than solely using local models while providing stronger data protection than fully relying on remote models. Furthermore, our method improves accuracy by 16.2% - 43.6% while reducing data leakage by 2.3% - 44.6% compared to existing data protection approaches.

Accurate Remaining Useful Life (RUL) estimation for CT X-ray tube filaments is critical to reduce unscheduled downtime and optimize maintenance in clinical settings. This work leverages full-lifecycle operational logs collected via an Internet of Medical Things (IoMT) platform and selects the parameter-stable Scout View mode to build physically consistent degradation sequences. A multi-stage preprocessing pipeline—anomaly removal, failure-threshold truncation, normalization, downsampling, and sliding-window segmentation — yields standardized multivariate inputs. To validate the effectiveness of our method, we conducted a comprehensive comparative analysis against five baseline paradigms: Hidden Markov Models (HMM), segmented linear regression, Gated Recurrent Units (GRU), 1D Convolutional Neural Networks (1D-CNN), and Transformers. Based on this, we develop a multivariate prediction framework using filament current, cumulative scan time, and cumulative usage days based on Long Short-Term Memory (LSTM) networks. To further improve temporal consistency, we introduce a novel historical path backtracking correction mechanism. Experiments on 38 complete filament life cycles (7 holdout test samples) demonstrate that the backtracking-corrected LSTM achieves superior performance, yielding a mean absolute percentage error (MAPE) of 9.55% and R² = 0.989, significantly outperforming all statistical and deep learning baselines. This IoMT-driven, deployable approach provides an effective solution for predictive maintenance, enabling proactive tube replacement planning and supporting resource optimization in smart hospital operations.

Kiran Ramnath, Kang Zhou, Sheng Guan, Soumya Smruti Mishra, Xuan Qi, Zhengyuan Shen, Shuai Wang, Sangmin Woo, Sullam Jeoung, Yawei Wang, Haozhu Wang, Han Ding, Yuzhe Lu, Zhichao Xu, Yun Zhou, Balasubramaniam Srinivasan, Qiaojing Yan, Yueyan Chen, Haibo Ding, Panpan Xu, Lin Lee Cheong. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Vision-Language Models (VLMs) have demonstrated impressive performance across a versatile set of tasks. A key challenge in accelerating VLMs is storing and accessing the large Key-Value (KV) cache that encodes long visual contexts, such as images or videos. While existing KV cache compression methods are effective for Large Language Models (LLMs), directly migrating them to VLMs yields suboptimal accuracy and speedup. To bridge the gap, we propose VL-Cache, a novel KV cache compression recipe tailored for accelerating VLM inference. In this paper, we first investigate the unique sparsity pattern of VLM attention by distinguishing visual and text tokens in prefill and decoding phases. Based on these observations, we introduce a layer-adaptive sparsity-aware cache budget allocation method that effectively distributes the limited cache budget across different layers, further reducing KV cache size without compromising accuracy. Additionally, we develop a modality-aware token scoring policy to better evaluate the token importance. Empirical results on multiple benchmark datasets demonstrate that retaining only 10% of KV cache achieves accuracy comparable to that with full cache. In a speed benchmark, our method accelerates end-to-end latency of generating 100 tokens by up to 2.33x and speeds up decoding by up to 7.08x, while reducing the memory footprint of KV cache in GPU by 90%.

Measuring similarities between different tasks is critical in a broad spectrum of machine learning problems, including transfer, multi-task, continual, and meta-learning. Most current approaches to measuring task similarities are architecture-dependent: (1) relying on pre-trained models, or (2) training networks on tasks and using forward transfer as a proxy for task similarity. In this paper, we leverage the optimal transport theory and define a novel task embedding for supervised classification that is model-agnostic, training-free, and capable of handling (partially) disjoint label sets. In short, given a dataset with ground-truth labels, we perform a label embedding through multi-dimensional scaling and concatenate dataset samples with their corresponding label embeddings. Then, we define the distance between two datasets as the 2-Wasserstein distance between their updated samples. Lastly, we leverage the 2-Wasserstein embedding framework to embed tasks into a vector space in which the Euclidean distance between the embedded points approximates the proposed 2-Wasserstein distance between tasks. We show that the proposed embedding leads to a significantly faster comparison of tasks compared to related approaches like the Optimal Transport Dataset Distance (OTDD). Furthermore, we demonstrate the effectiveness of our embedding through various numerical experiments and show statistically significant correlations between our proposed distance and the forward and backward transfer among tasks on a wide variety of image recognition datasets.

Learning from set-structured data is an essential problem with many applications in machine learning and computer vision. This paper focuses on a non-parametric, data-independent, and efficient learning algorithm from setstructured data using optimal transport and approximate nearest neighbor (ANN) solutions, particularly localitysensitive hashing. We consider the problem of set retrieval from an input set query. This retrieval problem requires 1) an efficient mechanism to calculate the distances/dissimilarities between sets and 2) an appropriate data structure for a fast nearest-neighbor search. To that end, we propose to use Sliced-Wasserstein embedding as a computationally efficient "set-2-vector" operator that enables downstream ANN with theoretical guarantees. The set elements are treated as samples from an unknown underlying distribution, and the Sliced-Wasserstein distance is used to compare sets. We demonstrate the effectiveness of our algorithm, denoted as Set Locality Sensitive Hashing (SLoSH), on various set retrieval datasets and compare our proposed embedding with standard set embedding approaches, including Generalized Mean (GeM) embedding/pooling, Featurewise Sort Pooling (FSPool), Covariance Pooling, and Wasserstein embedding and show consistent improvement in retrieval results, both in terms of accuracy and computational efficiency.

Out-of-distribution (OOD) data poses serious challenges in deployed machine learning models, so methods of predicting a model's performance on OOD data without labels are important for machine learning safety. While a number of methods have been proposed by prior work, they often underestimate the actual error, sometimes by a large margin, which greatly impacts their applicability to real tasks. In this work, we identify *pseudo-label shift*, or the difference between the predicted and true OOD label distributions, as a key indicator of this underestimation. Based on this observation, we introduce a novel method for estimating model performance by leveraging optimal transport theory, Confidence Optimal Transport (COT), and show that it provably provides more robust error estimates in the presence of pseudo-label shift. Additionally, we introduce an empirically-motivated variant of COT, Confidence Optimal Transport with Thresholding (COTT), which applies thresholding to the individual transport costs and further improves the accuracy of COT's error estimates. We evaluate COT and COTT on a variety of standard benchmarks that induce various types of distribution shift -- synthetic, novel subpopulation, and natural -- and show that our approaches significantly outperform existing state-of-the-art methods with up to 3x lower prediction errors.

Box representation has been extensively used for object detection in computer vision. Such representation is efficacious but not necessarily optimized for biomedical objects (e.g., glomeruli), which play an essential role in renal pathology. In this paper, we propose a simple circle representation for medical object detection and introduce CircleNet, an anchor-free detection framework. Compared with the conventional bounding box representation, the proposed bounding circle representation innovates in three-fold: (1) it is optimized for ball-shaped biomedical objects; (2) The circle representation reduced the degree of freedom compared with box representation; (3) It is naturally more rotation invariant. When detecting glomeruli and nuclei on pathological images, the proposed circle representation achieved superior detection performance and be more rotation-invariant, compared with the bounding box. The code has been made publicly available: https://github.com/hrlblab/CircleNet.

Abstract We present an approach for compressing volumetric scalar fields using implicit neural representations. Our approach represents a scalar field as a learned function, wherein a neural network maps a point in the domain to an output scalar value. By setting the number of weights of the neural network to be smaller than the input size, we achieve compressed representations of scalar fields, thus framing compression as a type of function approximation. Combined with carefully quantizing network weights, we show that this approach yields highly compact representations that outperform state‐of‐the‐art volume compression approaches. The conceptual simplicity of our approach enables a number of benefits, such as support for time‐varying scalar fields, optimizing to preserve spatial gradients, and random‐access field evaluation. We study the impact of network design choices on compression performance, highlighting how simple network architectures are effective for a broad range of volumes.

There has been a long pursuit for precise and reproducible glomerular quantification on renal pathology to leverage both research and practice. When digitizing the biopsy tissue samples using whole slide imaging (WSI), a set of serial sections from the same tissue can be acquired as a stack of images, similar to frames in a video. In radiology, the stack of images (e.g., computed tomography) are naturally used to provide 3D context for organs, tissues, and tumors. In pathology, it is appealing to do a similar 3D assessment. However, the 3D identification and association of large-scale glomeruli on renal pathology is challenging due to large tissue deformation, missing tissues, and artifacts from WSI. In this paper, we propose a novel Multi-object Association for Pathology in 3D (Map3D) method for automatically identifying and associating large-scale cross-sections of 3D objects from routine serial sectioning and WSI. The innovations of the Multi-Object Association for Pathology in 3D (Map3D) method are three-fold: (1) the large-scale glomerular association is formed as a new multi-object tracking (MOT) perspective; (2) the quality-aware whole series registration is proposed to not only provide affinity estimation but also offer automatic kidney-wise quality assurance (QA) for registration; (3) a dual-path association method is proposed to tackle the large deformation, missing tissues, and artifacts during tracking. To the best of our knowledge, the Map3D method is the first approach that enables automatic and large-scale glomerular association across 3D serial sectioning using WSI. Our proposed method Map3D achieved MOTA = 44.6, which is 12.1% higher than the non-deep learning benchmarks.