Pistachio: Advanced Synthetic Benchmark for Video Anomaly Detection
Discover Pistachio, a synthetic, balanced, and long-form video anomaly benchmark enhancing detection and understanding of complex video anomalies.
DeCo: Efficient Frequency-Decoupled Pixel Diffusion for AI Images
Discover DeCo, a novel frequency-decoupled pixel diffusion method boosting image generation speed and quality with superior FID scores on ImageNet.
Data-Efficient Text-to-Video Generation with Sparse Data
Discover data-efficient methods for controllable text-to-video generation using sparse synthetic data to achieve superior results with less training data.
SpecQuant: Ultra-Low-Bit Quantization for Large Language Models
Discover SpecQuant's spectral decomposition and adaptive truncation for efficient ultra-low-bit quantization in LLMs, boosting speed and reducing memory us...
Evaluating Speech Translation with Source-Aware Neural Metrics
Discover advanced source-aware neural MT metrics for accurate speech translation evaluation using ASR transcripts and back-translations.