Splattered Reality: How Easily Can 3D Gaussian Splatting Be Fooled?
![Even imperceptible adversarial perturbations, generated through an attack like [latex]PGD[/latex], demonstrably degrade the reconstruction quality of a feed-forward 3D Gaussian Splatting model-specifically, the NoPoSplat implementation on the RE10K dataset-as evidenced by diminished performance metrics like normalized Peak Signal-to-Noise Ratio (PSNR) and qualitative visual artifacts in newly rendered views, despite utilizing multiple reference views.](https://arxiv.org/html/2603.23686v1/x1.png)
A new study reveals that even subtle, carefully crafted disturbances can significantly degrade the quality of reconstructions from feed-forward 3D Gaussian Splatting models.
![Even imperceptible adversarial perturbations, generated through an attack like [latex]PGD[/latex], demonstrably degrade the reconstruction quality of a feed-forward 3D Gaussian Splatting model-specifically, the NoPoSplat implementation on the RE10K dataset-as evidenced by diminished performance metrics like normalized Peak Signal-to-Noise Ratio (PSNR) and qualitative visual artifacts in newly rendered views, despite utilizing multiple reference views.](https://arxiv.org/html/2603.23686v1/x1.png)
A new study reveals that even subtle, carefully crafted disturbances can significantly degrade the quality of reconstructions from feed-forward 3D Gaussian Splatting models.

A new study reveals that improving data retrieval in AI systems doesn’t guarantee more accurate responses, especially when dealing with complex policy questions.
New research reveals that training ReLU neural networks can be reframed as a convex optimization problem, unlocking theoretical benefits and improved training stability.

New research shows that combining artificially generated data with genetic insights significantly improves the accuracy of infectious disease predictions.

A comprehensive study reveals the vast scale of Telegram bots and their growing role in facilitating fraud, data leaks, and other cybercrimes.
A new review explores how deep learning is transforming our ability to understand animal movement and habitat preferences.
![For a population subjected to the GARP 50K challenge-assessed at a difficulty level of [latex]H=10[/latex]-a consumer-level cognitive-emotional intelligence index (CCEI) serves as a clear differentiator, with those achieving perfect scores ([latex]CCEI=1[/latex]) demonstrably exceeding the fitness threshold required for success.](https://arxiv.org/html/2603.23993v1/x3.png)
New research shows that incorporating principles of revealed preference – even with imperfect consumer behavior – can significantly improve the accuracy of time-series forecasting models.
Inspired by the principles of free-market economics, researchers have developed a novel optimization framework capable of discovering solutions in challenging and open-ended domains.

Researchers have developed a novel graph neural network that enhances the detection of money laundering by better analyzing transaction patterns.

Researchers have developed a novel framework that leverages dynamic relationships between stocks to improve the accuracy of trend forecasting.