arXiv:2601.08026v1 Announce Type: new Abstract: Scientific compound figures combine multiple labeled panels into a single image, but captions in real pipelines are often missing or only provide figure-level summaries, making panel-level understanding difficult. In this paper, we propose FigEx2, visual-conditioned framework that localizes panels and generates panel-wise captions directly from the compound figure. To mitigate the impact of diverse phrasing in open-ended captioning, we introduce a noise-aware gated fusion module that adaptively filters token-level features to stabilize the […]

Ver mais

Prompt Chaining: When One Prompt Isn’t Enough If you’ve ever tried to squeeze an entire project into one prompt—requirements → solution → plan → risks → final doc—you already know how it ends: it skips steps, it forgets constraints, it gives you a “confident” answer you can’t easily verify, and the moment something is wrong, you have no idea where the mistake happened. Prompt Chaining is the fix. Think of it as building a workflow where each prompt […]

Ver mais

Innovator-VL argues scale isn’t destiny: with ~5M curated examples, native-resolution vision tokens, and RL-for-reasoning, it matches bigger models—reproducibly.

Ver mais

arXiv:2601.12150v1 Announce Type: new Abstract: Despite their prominent performance on tasks such as ROI classification and segmentation, many pathology foundation models remain constrained by a specific input size e.g. 224 x 224, creating substantial inefficiencies when applied to whole-slide images (WSIs), which span thousands of resolutions. A naive strategy is to either enlarge inputs or downsample the WSIs. However, enlarging inputs results in prohibitive GPU memory consumption, while downsampling alters the microns-per-pixel resolution and obscures critical morphological details. […]

Ver mais

Deep learning models for radiology interpretation increasingly rely on multi-institutional data, yet privacy regulations and distribution shift across hospitals limit central data pooling. Federated learning (FL) allows hospitals to collaboratively train models without sharing raw images, but current FL algorithms typically assume a static data distribution. In practice, hospitals experience continual evolution in case mix, annotation protocols, and imaging devices, which leads to catastrophic forgetting when models are updated sequentially. Federated continual learning (FCL) aims to reconcile these […]

Ver mais

arXiv:2601.02391v1 Announce Type: new Abstract: Wearable devices such as AI glasses are transforming voice assistants into always-available, hands-free collaborators that integrate seamlessly with daily life, but they also introduce challenges like egocentric audio affected by motion and noise, rapid micro-interactions, and the need to distinguish device-directed speech from background conversations. Existing benchmarks largely overlook these complexities, focusing instead on clean or generic conversational audio. To bridge this gap, we present WearVox, the first benchmark designed to rigorously evaluate […]

Ver mais

Intrusion detection aims to effectively detect abnormal attacks in the Internet of Things (IoT) network, which is crucial for cybersecurity. However, traditional intrusion detection methods are difficult to effectively extract data features from traffic data, and most existing models are too complex to be deployed on edge servers. Addressing this need, this paper proposes a hybrid feature selection method and a lightweight deep learning intrusion detection model. Firstly, the data features space is reduced by using variance filtering, […]

Ver mais

Hi everyone, I’ve been working in NLP for several years, and my role has gradually shifted from training models to mainly using LLM wrappers. I’m concerned that this kind of work may become less in demand in the coming years. I now have an opportunity to transition into Computer Vision. After about two months of self-study and research, I feel that the gap between academic research and real-world applications in CV is relatively large, and that the field […]

Ver mais

arXiv:2601.17063v1 Announce Type: new Abstract: Recently, Mixture-of-Experts (MoE) models have gained attention for efficiently scaling large language models. Although these models are extremely large, their sparse activation enables inference to be performed by accessing only a fraction of the model at a time. This property opens the possibility of on-device inference of MoE, which was previously considered infeasible for such large models. Consequently, various systems have been proposed to leverage this sparsity and enable efficient MoE inference for […]

Ver mais

Memory is increasingly central to Large Language Model (LLM) agents operating beyond a single context window, yet most existing systems rely on offline, query-agnostic memory construction that can be inefficient and may discard query-critical information. Although runtime memory utilization is a natural alternative, prior work often incurs substantial overhead and offers limited explicit control over the performance-cost trade-off. In this work, we present textbf{BudgetMem}, a runtime agent memory framework for explicit, query-aware performance-cost control. BudgetMem structures memory processing […]

Ver mais