All papers

Presents a GPU nearest-neighbor search method that uses half-precision certified exclusion to deliver exact visual retrieval efficiently.

Proposes invertible bi-Lipschitz adapters to align pretrained latent spaces with semantic representations for flow matching without retraining encoders.

Presents an auditable framework that uses risk budgeting to selectively deploy generative enhancers over base restoration models in low-light images.

Introduces a label-efficient self-supervised learning framework that splits volumetric CT scans into local slice and global context representations.

Proposes a script-agnostic one-shot font generation framework separating style and content representations across multiple spatial scales.

Presents a backpropagation-free test-time adaptation method that purges unreliable 3D point cloud tokens without requiring candidate size search.

Presents a keypoint descriptor method that leverages global line correspondences to improve feature matching in repetitive or textureless scenes.

Proposes a framework that increases image diffusion throughput by fusing and processing multiple generation samples simultaneously in manifold space.

Introduces a benchmark and dataset of over two million images to evaluate deepfake detection models on partial-face manipulations.

Presents a multi-modal registration framework for longitudinal whole-body PET/CT that preserves metabolic tumor quantification metrics.

Presents an agentic framework that iteratively refines 3D scene geometry from multi-view video captures against standing visual observations.

Combines lightweight learned guidance with classical filter banks for efficient, low-latency video denoising in communication pipelines.

Introduces an evaluation protocol for medical image segmentation uncertainty metrics based on clinical deferral policies and fixed review budgets.

Presents a feed-forward 3D Gaussian splatting method that uses entropy-guided 3D anchors instead of pixel-aligned predictions for efficient reconstruction.

Proposes a method to dynamically adjust stochastic spread during inference to improve the reliability of weather ensemble forecasts.

Presents a training-free token merging approach combining spatial consistency and feature similarity to speed up vision transformers in segmentation.

Releases an extended 50-year multispectral satellite benchmark for typhoon intensity regression and analyzes systematic model biases.

Presents a clinical workflow using intraoperative fluoroscopy to provide subsurface ground truth for validating markerless spinal tracking systems.

Proposes an adaptive multi-sequence feature difference module to synthesize contrast-enhanced MRI scans from non-contrast imaging sequences.

Validates an unsupervised domain adaptation model to classify normal and abnormal retinal development across diverse optical coherence tomography systems.

Introduces a dormant spiking neuron mechanism that skips subthreshold computations to reduce computational overhead in spiking neural networks.

Introduces a parameter-efficient multimodal fusion framework inside frozen vision backbones that handles missing modalities during semantic segmentation.

Introduces an open-source Python toolbox combining computer vision and speech models to analyze multimodal human nonverbal communication signals.

Proposes a single-stage joint diffusion method to forecast solar irradiance directly from sky images without requiring a separate regression model.

Refines DINO foundation features using self-supervised low-rank adaptation to improve partial 3D shape matching.

The paper benchmarks the performance degradation of medical foundation models when transferring from adult chest X-rays to pediatric patients.

The paper applies implicit neural representations with temporal consistency constraints to register periodic medical image sequences across time.

The paper presents a dual-frame coordinate transformer to generate 3D assets with separable, editable component parts.

The paper reconstructs relightable 3D garment meshes from 2D sketches annotated with sparse lighting symbols.

The paper introduces a graph-structured relational reward shaping framework for diffusion-based trajectory planning in autonomous driving.

The paper presents a multimodal vision transformer framework combining thermal guidance and RGB features for image segmentation in low-light environments.

The paper introduces a sparse convolutional masked autoencoder framework to improve data-efficient pretraining on radar satellite imagery.

The paper trains discrete image tokenizers by reconstructing pretrained self-supervised feature embeddings rather than raw pixel values.

The paper tests whether pathology foundation models can retrieve relevant historical cases for rare salivary gland tumors without task-specific fine-tuning.

The paper introduces a multiple instance learning method to recover coexisting secondary lesions from whole-slide pathology images labeled only by dominant severity.

The paper develops an FPGA-based vision system that enforces explicit latency bounds for autonomous vehicle 3D occupancy prediction.

The paper presents a refiner that models motion proposals as heterogeneous temporal graphs to improve trajectory forecasting for autonomous vehicles.

The paper uses wavelet subband decomposition with language models to improve entropy modeling for lossless image compression.

Analyzes image denoisers on repetitive textures and finds fine-tuning on textures leads to catastrophic forgetting of generic images.

Presents an annotation-free retinal vessel segmentation pipeline combining classical filtering, Bayesian deep learning, and graph reconstruction.

Maximizes batch Gram matrix determinants during initial diffusion noise optimization to improve overall sample diversity in image generation.

Combines vision transformer features with color histograms to cluster public transit vehicles in surveillance video without manual labels.

Proposes an online reinforcement learning approach that uses attention routing as stochastic exploration to align visual diffusion generators with rewards.

Presents a real-time tracking-by-detection method that dynamically scales search regions based on predictive motion uncertainty.

Introduces a feed-forward model that predicts dense 4D point correspondences across articulated human bodies from monocular RGB images.

Introduces an auditing framework that checks counterfactual execution traces to ensure language model routing uses only legitimate information.

Proposes a hybrid architecture combining global state-space modeling with local transformer attention and lexical memory for efficient inference.

Presents a benchmark with open-ended tasks and rubrics to assess social awareness and response generation in Chinese.

Studies the transferability of internal activation interventions to reduce unsafe tool execution across model precisions and languages.

Introduces an off-policy reinforcement learning algorithm for language models using verifiable rewards without importance sampling weights.

Introduces a benchmark combining natural, scripted, and synthetic audio to evaluate phonetic error detection in speech models.

Tests whether linguistically accurate morpheme boundaries in tokenizers improve translation performance compared to scrambled control segmentations.

Introduces a method to filter unsupported compressor inferences from long-horizon agent summaries to avoid compounding execution errors.

Deconstructs multiple-choice option bias in language models into two factors tracking answer flips and positional skew.

Explores methods to extract the underlying decisive sub-questions that drive language model decisions across diverse generated outputs.

Proposes an adaptive reasoning framework that supervises intermediate retrieval states and stops reasoning steps dynamically to cut latency.

Introduces a benchmark to test model generalization under systematic convenience sampling biases in tabular data.

The paper proposes SHIP, an architecture for extracting structured differentially private representations from text for analytical queries.

The paper proposes RoPERelay to reorganize positional geometry and rotary embeddings after sparse KV cache token eviction.

The paper introduces a causal intervention framework to distill transferable self-generated in-context learning demonstrations.