All papers

The paper proposes a generative quantum machine learning method to create synthetic breast ultrasound images for training cancer detection models.

The paper presents a block-wise streaming geometric foundation model for continuous 3D reconstruction and simulation in autonomous driving.

The paper introduces a visual-inertial SLAM system that applies square-root bundle adjustment to achieve faster multi-camera tracking on low-power devices.

The paper proposes an evaluation metric using a texture-trained GAN discriminator backbone to measure texture realism in generated images.

The paper introduces a benchmark evaluating whether coding agents can construct interactive, code-based 3D simulations that capture underlying physical principles.

The paper presents a two-stage multiscale residual convolutional network to segment and quantify carotid artery atherosclerotic inflammation on PET images.

The paper introduces an efficient method that couples generative complete-geometry prediction with active shape models for category-level 6D pose estimation.

The paper proposes a few-step distillation framework for native 3D generation to mitigate sampling-induced latent drift that damages texture synthesis.

The paper presents a predictive KV cache quantization method exploiting temporal predictability to reduce memory overhead in autoregressive video generation models.

The paper presents a four-gate verification protocol to evaluate whether audio-conditioned visual prediction models genuinely rely on input audio signals.

The paper adapts the Segment Anything Model for automated brain tumor segmentation using multi-scale convolutional experts without requiring manual inference prompts.

The paper evaluates fine-tuned video transformers on an expanded video dataset to detect instances of sexual harassment for surveillance applications.

The paper introduces a method that unifies 3D reconstruction and generation within a shared anchor-guided coordinate space to preserve observed geometry.

The paper proposes an agent harness that autonomously evolves questioner and solver workflows from unlabeled video data without updating base model parameters.

The paper proposes a unified latent diffusion architecture with a custom variational autoencoder for multimodal remote sensing image generation.

The paper introduces a lightweight test-time attention framework that leverages single-photon sensor noise as an informative prior for indoor depth estimation.

The paper introduces a flow-free deep network for smartphone document image unwarping using geometric consistency and undistorted document supervision.

The paper introduces a benchmark to evaluate text-to-video models across controlled emotional expression levels and specific audiovisual affective cues.

The paper develops an automated method that measures semantic saliency by evaluating how removing individual objects alters multimodal model scene descriptions.

The paper introduces structured stateful relational scripts and a paired dataset to coordinate persistent identities and audio events in multi-shot video generation.

The paper proposes an adaptive image tokenizer that balances pixel reconstruction and semantic features to prevent discarding critical semantic content during compression.

The paper introduces a 3D Gaussian neural field transformer for efficient, camera-based occupancy prediction in robotic perception systems.

The paper proposes a training-free 3D editing framework that estimates edit regions without masks to preserve unedited source geometry and appearance.

The paper presents a simulator-grounded benchmark to evaluate whether image-to-video generation models capture causal physical cascades and counterfactual dynamics faithfully.

The paper introduces an image augmentation pipeline that generates diverse visual variants to anchor multimodal chain-of-thought reasoning in visual evidence.

The paper presents a theoretical framework using nesting properties to calculate distance bounds and compare language model truncation sampling algorithms efficiently.

The paper introduces an algorithm that fits multiple ellipsoids in multi-dimensional data by leveraging tangent equations rather than raw points alone.

The paper presents a framework aligning visual image restoration with navigation decisions for vision-language navigation in degraded indoor fire environments.

This paper introduces a lightweight detection framework combining early channel recalibration and multiscale fusion for real-time traffic sign identification.

This paper introduces a dual-branch framework combining mask-level and image-level supervision for open-vocabulary mammography lesion segmentation.

This paper accelerates world action models by compressing multi-step action denoising into a single forward step parallelized with visual prefilling.

This paper proposes predicting future instrument skeletons from planned actions to guide visually and structurally consistent surgical video generation.

This paper introduces a sparse connector allowing cross-layer feature sharing across different foundation models for domain-generalized semantic segmentation.

This paper introduces prototype-purity pseudo-labeling to extract accurate instance-level annotations from weakly supervised whole-slide pathology images.

This paper proposes offline reinforcement learning conditioned on quantization levels to recover task completion performance in quantized vision-language-action policies.

This paper introduces a benchmark evaluating whether vision-language models can infer underlying physical laws from motion videos and generate runnable simulation code.

This paper introduces a question-agnostic adversarial text-watermarking method that severely degrades document understanding in vision-language models.

This paper proposes estimating 3D object poses within partial scenes by predicting visibility fields in canonical object space.

This paper introduces a training-free prototype memory readout to prevent out-of-vocabulary overconfidence and silent hallucinations in remote sensing classification.

This paper presents a spiking mixture-of-experts architecture to improve the resilience of multimodal 3D occupancy perception against missing sensor inputs.

This paper demonstrates that accurate registration can cause temporal consistency losses to freeze history accumulation and degrade real-time supersampling accuracy.

This paper introduces a competitive bi-modal slot attention method to cleanly segment distinct sounding objects in multi-source audio-visual scenes.

This paper presents a framework to reconstruct 4D hand-object interactions from human demonstration videos to supervise robot manipulation policies.

This paper introduces a recursive self-improvement method that unifies instruction-following and instruction generation within a shared model for vision-language navigation.

This paper presents a depth-conditioned video recovery framework that selects a single reliable RGB reference to restore damaged video bitstreams efficiently.

This paper proposes an object-centric 3D memory framework to maintain persistent object representations and world consistency during video generation.

This paper evaluates visual test-time training for remote sensing change detection using a Siamese encoder-decoder architecture with adaptive fast weights.

This paper introduces a learnable probabilistic instance sampler to handle severe class imbalance and prioritize critical diagnoses in multilabel learning.

This paper presents an unsupervised test-time adaptation method that uses movement-induced prediction variance to adapt object recognition models in mobile agents.

This paper shows that wide viewpoint coverage produces significantly better 3D room reconstructions than selecting only sharp, motion-free camera frames.

This paper proposes a framework that jointly evolves visual evidence presentation and reading procedures to improve multimodal agent task performance.

This paper introduces a training-free caching method that separates cache synchronization and computation repair to accelerate autoregressive video diffusion generation.

This paper presents a verification and recovery framework using vision-language models to correct tracking errors and mask drift in video object segmentation.

Proposes a knowledge distillation method that focuses learning signals on specific structural disagreements between teacher and student representations.

Introduces a corpus of vision-language model generated importance scores to improve temporal sampling in video action recognition.

Implements an aggressively pruned spiking neural network to perform event-based visual place recognition on commodity CPUs in microseconds.

Presents an open-source framework using context forests and Viterbi dynamic programming to extract and number rib centerlines in CT scans.

Analyzes failure prediction and component selection methods to evaluate which semantic segmentation predictions to retain or inspect.

Develops an incremental long-video captioning method using a persistent identity gallery and evolving narrative summary.

Evaluates router-learning strategies for combining domain-specialized diffusion experts in multi-domain medical image generation.