Sheng Li

A temporal neural denoiser validated on stochastic 2D Gaussian Splatting rendering, combining dual-path exponential moving average accumulation, per-pixel learned trust prediction for history validation, a fixed anisotropic spatial filter and a variance-gated composition with stabilization is proposed.

Instead of forcing entire videos into a single, lossy feed-forward context, training an agent to actively read and write temporal memory yields 78.8% accuracy on MINERVA with virtually zero performance drop on long-duration clips.

GazeFS reduces gaze trajectory errors by up to 0.400 degrees while maintaining temporal smoothness, revolutionizing target-centered gaze interaction.

LLM refusals often prioritize explicitness over user sensitivity, risking harm in sensitive contexts where a more nuanced approach is critical.

DuplexGen achieves more authentic conversational dynamics by allowing timing to emerge organically rather than relying on rigid, handcrafted rules.

Cached LLM probability retrieval can improve ASR performance without the heavy lifting of retraining, outperforming traditional methods in a majority of tested scenarios.

A novel dataset and framework that not only detects AI-generated images but also provides coherent visual evidence and explanations, setting a new standard for authenticity assessment.

ThinkOmni achieves unprecedented cross-dataset generalization in audio forgery detection by explicitly linking forensic reasoning to manipulation cues.