3466 papers
To address the trade-off between backward compatibility and discriminative power in retrieval model updates, this paper proposes a method that relaxes alignment constraints by introducing Neighbor-Driven and Optimization-Driven perturbations to old feature prototypes, enabling the new model to align with a pseudo-old feature space while preserving its ability to distinguish closely clustered classes.
This paper proposes a zero-trust QKD relay architecture that utilizes fully homomorphic encryption to perform intermediate key re-encryption without exposing plaintext, thereby eliminating the security risks of trusted nodes while ensuring compatibility with existing ETSI standards and enhancing crypto-agility.
This paper proposes a unified framework for occlusion-robust two-hand reconstruction that combines a fusion-alignment encoder to implicitly integrate heterogeneous 2D structural priors from vision foundation models with a penetration-free diffusion model that guides 3D pose generation toward collision-free, kinematically coherent interactions.
This paper introduces MediTools, an AI-powered prototype application that leverages large language models to revolutionize medical education through interactive dermatology case simulations, enhanced literature analysis, and automated medical news summaries, while validating its potential through a survey of medical professionals and students.
This paper proposes a taxonomy and design framework for 6G Medium Access Control protocols that integrates push-based and pull-based communication schemes, leveraging AI-driven, goal-oriented criteria to optimize resource allocation and co-existence within Open-RAN architectures.
This paper proposes a point-based instance completion model that leverages cross-attention to integrate sparse scene constraints for robustly completing objects at arbitrary scales and poses, validated by a new dataset called ScanWCF that demonstrates superior completion quality and scene plausibility over existing methods.
This paper introduces Lipschitz continuity-constrained Ensemble Learning (LEL), a novel framework that enhances the stability, accuracy, and robustness of intra-subject EEG-based emotion recognition by enforcing Lipschitz constraints on Transformer components and utilizing a learnable ensemble fusion strategy, achieving superior performance on three public benchmark datasets.
This paper introduces GeoNav, a multi-modal agent that enhances MLLMs for long-range aerial navigation by employing a three-phase, coarse-to-fine reasoning process supported by dual-scale spatial representations (a global cognitive map and a local scene graph) and a spatial chain-of-thought mechanism, achieving state-of-the-art performance on the CityNav benchmark.
This paper proposes the Orthogonally-constrained Variational Information Bottleneck (O-VIB) framework, which leverages automatic relevance determination and orthogonality constraints to compress multi-view visual features for precise, bandwidth-efficient localization on resource-constrained mobile platforms at the network edge.
This paper compares Bayesian optimization, exhaustive search, genetic algorithms, and stochastic gradient descent for optimizing industrial robot base poses, finding that while all methods reduce cycle time, stochastic gradient descent achieves the highest success rate in real-world tasks and genetic algorithms yield the lowest final costs.
The paper proposes SFIBA, a spatial-based full-target invisible backdoor attack that ensures trigger specificity and stealthiness in black-box settings by restricting triggers to local spatial regions and employing a frequency-domain injection method, thereby achieving high attack performance while evading existing defenses.
This paper presents a holistic optimization framework that jointly determines the composition, base placement, and trajectory of modular robots to minimize cycle time, achieving up to a 25% reduction in task duration and significantly higher feasibility rates compared to composition-only optimization, with successful real-world deployment validated in nine out of ten cases.
This paper proposes and validates a real-time coordinated routing strategy using connected and automated vehicles (CAVs) in mixed-traffic environments to dynamically reroute them away from dedicated bus lanes, thereby simultaneously improving bus schedule adherence and CAV travel efficiency.
A randomized study with 94 novice counselors demonstrates that combining LLM-simulated practice with structured feedback significantly improves client-centered microskills and empathy, whereas practice alone fails to produce gains and may even lead to a decline in empathetic performance.
The paper introduces DeepSparse, a foundation model for sparse-view CBCT reconstruction that utilizes a novel DiCE network and a HyViP pretraining framework to achieve superior image quality with reduced radiation exposure while overcoming the computational and generalization limitations of existing methods.
HEXGEN-FLOW is a hierarchical scheduling framework designed to optimize agentic Text-to-SQL inference on heterogeneous GPU clusters, achieving significant improvements in tail latency and throughput over existing LLM serving systems by addressing multi-stage dependencies and strict latency requirements.
This paper introduces MCAD, a comprehensive multimodal benchmark featuring 204 sequences and 30,000 frames designed to advance generic cooperative autonomous driving research across multiple tasks, accompanied by baseline evaluations and a novel adaptive fusion method that balances communication efficiency with perception accuracy.
This paper presents FoldNet, a framework that addresses the challenge of robotic garment folding by generating a large-scale synthetic dataset through keypoint-driven asset and demonstration synthesis, and employs a keypoint-based KG-DAgger algorithm to train a robust closed-loop policy that achieves a 75% success rate in real-world tasks.
Diffusion-SAFE is a closed-loop shared autonomy framework that utilizes dual diffusion models to predict human intent and generate safety-guided driving plans, enabling smooth and continuous control handovers by adjusting a forward-diffusion ratio to balance human intent with safety constraints.
MTVCraft introduces a novel framework that tokenizes raw 3D motion sequences into 4D motion tokens and leverages a motion-aware Video DiT to enable robust, generalizable, and flexible character image animation, surpassing traditional methods that rely on 2D pose images.