New Speech Models Target Real-Time Phone Scam Detection
Three new papers tackle the challenge of catching telecom fraud from live audio, focusing on speed, adaptability, and structured reasoning.
Preprints and papers, rewritten in plain English.
Three new papers tackle the challenge of catching telecom fraud from live audio, focusing on speed, adaptability, and structured reasoning.
Four new arXiv papers propose distinct improvements to time-series forecasting, from leveraging physical knowledge to aligning training objectives.
Four arXiv papers push robot manipulation beyond rigid parts, addressing deformable dynamics, asset generation, cable shaping, and skill reuse without demonstrations.
Four arXiv preprints address distinct failure modes in policy learning, from experimental overfitting to visual domain shift.
A wave of arXiv papers addresses the practical barriers to deploying vision-language-action models in robotics, from inference speed and fine-tuning costs to the need for more training data.
New arXiv preprints address four distinct failure modes of 3D Gaussian Splatting: reflective surfaces, material decomposition, redundant Gaussians, and streaming RGB-D input.