17 accepted papers
Traditional knowledge distillation relies on simple MSE or KL divergence losses that fail to capture the complex distributional relationships between teacher and student model representations. We propose FlowDistill, a novel distillation framework that employs normalizing flows to model and transfer
Unlearning in large language models (LLMs) is critical for regulatory compliance and for building ethical generative AI systems that avoid producing private, toxic, illegal, or copyrighted content. Despite rapid progress, in this work we show that \textit{almost all} existing unlearning methods fail…
Large reasoning models achieve strong performance through test-time scaling, but this incurs substantial computational overhead due to long decoding from short prompts. While sparse attention can reduce latency and memory usage, existing methods often degrade reasoning accuracy because selection err…
The growing demand for efficient image segmentation in marine ecological studies is currently constrained by two key factors: the high computational requirements of models such as the Segment Anything Model (SAM) and the degraded accuracy of lightweight models in underwater environments. To overcome
Recent vision language models (VLMs) have made remarkable strides in generative modeling with multimodal inputs, particularly text and images. However, their susceptibility to generating harmful content when exposed to unsafe queries raises critical safety concerns. While current alignment strategie…
Machine unlearning (MU) for large language models (LLMs), commonly referred to as LLM unlearning, seeks to remove specific undesirable data or knowledge from a trained model, while maintaining its performance on standard tasks. While unlearning plays a vital role in protecting data privacy, enforcin…
Foundation models for weather science are pre-trained on vast amounts of structured numerical data and outperform traditional weather forecasting systems. However, these models lack language-based reasoning capabilities, limiting their utility in interactive scientific workflows. Large language mode…
In recent years, Wi-Fi sensing has attracted much attention due to the widespread deployment of communication devices. Due to advancements in signal processing algorithms, contactless sensing technology based on Wi-Fi signals has now been widely applied. However, the limited bandwidth of Wi-Fi syste…
Multi-turn interactions between large language models (LLMs) and users naturally include implicit feedback signals. If an LLM responds in an unexpected way to an instruction, the user is likely to signal it by rephrasing the request, expressing frustration, or pivoting to an alternative task. Such s…
Input saliency aims to quantify the influence of input tokens on the output of large language models (LLMs), which has been widely used for prompt engineering, model interpretability, and behavior attribution. Despite the proliferation of saliency techniques, the field lacks a standardized and rigor…
With the ability to manipulate objects contactlessly, acoustic levitation has excellent potential in the micro-assembly industry. Fixed ultrasonic phased arrays (UPA) widely used in acoustic levitation control objects with restricted space and inflexibility. Here, an acoustic array robot based on th
Current language models demonstrate remarkable proficiency in text generation. However, for many applications it is desirable to control attributes, such as sentiment, or toxicity, of the generated language—ideally tailored towards each specific use case and target audience. For auto-regressive lang…
Test time augmentation (TTA) has been a promising tool for improving the robustness against out-of-distribution data at inference time. Recent TTA methods try to learn predictive transformations which are supposed to provide the best performance gain on each test sample. However, existing methods ar…
"Recent foundation models trained on a tremendous scale of data have shown great promise in a wide range of computer vision tasks and application domains. However, less attention has been paid to the marine realms, which in contrast cover the majority of our blue planet. The scarcity of labeled data…
Visual object tracking (VOT) is one of the most fundamental tasks in computer vision community. State-of-the-art VOT trackers extract positive and negative examples that are used to guide the tracker to distinguish the object from the background. In this paper, we show that this characteristic can b…
Single-Photon Image Super-Resolution (SPISR) aims to recover a high-resolution volumetric photon counting cube from a noisy low-resolution one by computational imaging algorithms. In real-world scenarios, pairs of training samples are often expensive or impossible to obtain. By extending Equivariant…
Intersection over union (IoU) score, also named Jaccard Index, is one of the most fundamental evaluation methods in machine learning. The original IoU computation cannot provide non-zero gradients and thus cannot be directly optimized by nowadays deep learning methods. Several recent works generaliz…