LiT (locked-image text tuning)
LiT, or locked-image text tuning, is a method in machine learning that enhances text generation by integrating visual information from images.
Technology & Innovation
Role prompting
Role prompting is a technique used in artificial intelligence and natural language processing where a system is given a specific role or persona to guide its responses. This method helps shape the behavior and output of AI models by instructing them to adopt particular perspectives or functions.
Model compression
Model compression refers to a set of techniques aimed at reducing the size and computational requirements of machine learning models while maintaining their performance. It enables deployment of models on resource-constrained devices and improves inference efficiency.
Zero-shot prompting
Zero-shot prompting is a technique in natural language processing where a model is given instructions or queries without prior examples, enabling it to perform tasks it was not explicitly trained on. This approach leverages the generalization capabilities of large language models to interpret and respond to new prompts directly.
Waymo Simulator
Waymo Simulator is a tool developed by Waymo to test and validate autonomous driving software in a controlled virtual environment.
Opus (audio codec with neural enhancement)
Opus with neural enhancement is an advanced audio codec that integrates conventional Opus compression with neural network-based processing to improve audio quality, especially at low bitrates. This hybrid approach aims to deliver clearer, more natural sound for applications in streaming, communication, and media playback.
BART (language model)
BART (Bidirectional and Auto-Regressive Transformer) is a neural network-based sequence-to-sequence language model developed by Facebook AI. It combines bidirectional and autoregressive transformers to enhance text generation and understanding tasks.
Representation learning
Representation learning is a set of techniques in machine learning that enable systems to automatically discover and extract useful features or representations from raw data. It plays a crucial role in improving the performance of algorithms by transforming data into formats that are easier to analyze and interpret.
Judea Pearl
Judea Pearl is a prominent computer scientist known for his contributions to artificial intelligence and causal inference.
UCF101 (action recognition dataset)
UCF101 is a widely used dataset for action recognition in videos, containing 13,320 clips across 101 action categories.