7576 papers
The paper introduces SpecTemp, a reinforcement learning-based framework that enhances the efficiency of long video understanding by decoupling temporal perception and reasoning through a cooperative dual-model design, where a lightweight draft MLLM proposes salient frames for verification by a powerful target MLLM, thereby significantly accelerating inference while maintaining competitive accuracy.
The paper proposes TARDis, a novel physics-aware framework that addresses incomplete multi-modal tumor segmentation by disentangling time-invariant anatomical features from time-dependent hemodynamic dynamics, enabling robust diagnosis even when temporal CT phases are missing due to clinical constraints.
This paper introduces a self-supervised AI-generated image detection framework that leverages EXIF metadata to learn intrinsic photographic features, achieving state-of-the-art generalization and robustness across diverse generative models through one-class and binary detection strategies.
This paper introduces FRIEDA, a rigorous benchmark for evaluating multi-step cartographic reasoning in large vision-language models, revealing a significant performance gap between current state-of-the-art systems and human capabilities in interpreting complex spatial relationships across map images.
SHARP is a novel, real-time monocular view synthesis method that uses a single feedforward neural network pass to regress metric 3D Gaussian parameters from a single image, achieving state-of-the-art photorealistic rendering with significantly reduced synthesis time and superior zero-shot generalization compared to prior models.
This paper proposes a novel event-based 3D Gaussian Splatting framework that decouples geometry and radiance rendering into event-by-event and snapshot-based branches, respectively, to achieve state-of-the-art, prior-free 3D reconstruction with high temporal resolution and sharp edge details.
ColaVLA is a unified vision-language-action framework that addresses the latency and modality mismatch of existing VLM-based planners by transferring cognitive reasoning into a compact latent space and employing a hierarchical parallel decoder to achieve state-of-the-art, efficient, and safe trajectory planning on the nuScenes benchmark.
This paper introduces WMReward, an inference-time alignment method that leverages a latent world model (VJEPA-2) as a reward signal to steer video generation trajectories, significantly improving physics plausibility across various conditioning settings and securing first place in the ICCV 2025 Perception Test PhysicsIQ Challenge.
CPiRi is a novel framework for multivariate time series forecasting that combines a spatio-temporal decoupling architecture with permutation-invariant regularization to overcome the limitations of existing channel-dependent and independent models, achieving state-of-the-art performance, robustness to channel reordering, and strong inductive generalization to unseen channels.
This paper proposes a learning-based HDR restoration framework for modulo imaging that combines scale-equivariant regularization with a feature lifting input design to effectively distinguish natural edges from wrapping artifacts and achieve state-of-the-art reconstruction performance.
This paper introduces CityGenAgent, a natural language-driven framework that utilizes a two-stage learning strategy of Supervised Fine-Tuning and Reinforcement Learning to hierarchically generate high-quality, editable, and semantically aligned 3D cities through interpretable Block and Building programs.
The paper introduces Erase at the Core (EC), a model-agnostic framework that enforces comprehensive machine unlearning by applying multi-layer contrastive learning and deep supervision to eliminate superficial forgetting and substantially reduce representational similarity across the entire network hierarchy while preserving performance on retained data.
PixelRush is a novel, training-free framework that achieves ultra-fast, high-resolution image generation by enabling efficient patch-based denoising in a single step, reducing 4K image creation time from minutes to approximately 20 seconds while maintaining superior visual quality.
This paper introduces MAILS, a novel map-free LiDAR relocalization framework specifically designed for UAVs that leverages locality-preserving attention and coordinate-independent mechanisms to achieve high-precision positioning in GNSS-denied environments, accompanied by the creation of a new large-scale dataset to address the lack of realistic UAV flight data.
COOPERTRIM is an adaptive data selection framework for cooperative perception that leverages temporal continuity and a novel conformal uncertainty metric to dynamically filter redundant information, achieving significant bandwidth reduction (up to 80%) while maintaining or improving detection and segmentation accuracy compared to existing methods.
Diff-Aid is a lightweight, plug-and-play inference-time method that adaptively modulates text-image interactions across transformer blocks and denoising timesteps to enhance prompt adherence and visual quality in rectified text-to-image generation while providing interpretable insights into the alignment process.
SceneTok introduces a novel tokenizer that compresses 3D scene view sets into a small, permutation-invariant set of unstructured tokens, enabling state-of-the-art reconstruction, flexible novel view synthesis, and efficient scene generation with significantly higher compression than existing methods.
This paper proposes a learning-free, prototype-guided framework for multimodal dataset distillation that leverages CLIP embeddings and an unCLIP decoder to synthesize images, thereby achieving state-of-the-art cross-architecture generalization without the computational costs and architectural limitations of existing optimization-based methods.
One2Scene is a novel framework that generates geometrically consistent, explorable 3D scenes from a single image by decomposing the task into panorama generation, 3D scaffold construction via multi-view stereo matching on sparse anchor views, and novel view synthesis, thereby overcoming the severe distortions and artifacts common in existing methods during large camera motions.
This paper reframes Test-Time Training (TTT) with KV binding not as a memorization-based online meta-learning process, but as a form of learned linear attention, a perspective that explains puzzling model behaviors and enables principled architectural simplifications and efficient parallel formulations.