12 terms
Showing all terms starting with T
A sampling parameter that controls the randomness of LLM outputs. Higher values produce more creative responses; lower values are more deterministic.
AI that generates visual images from natural language text descriptions, as in Midjourney, DALL-E, and Stable Diffusion.
AI technology that converts written text into natural-sounding spoken audio, used in voice assistants, audiobooks, and accessibility tools.
The units of text (roughly 4 characters or 3/4 of a word in English) that LLMs process. Costs and context limits are measured in tokens.
Using knowledge gained from training on one task or domain and applying it to a different but related task, reducing the need for large datasets.
The neural network architecture behind most modern LLMs, using attention mechanisms to process sequential data with high parallelism.
An NLP task that assigns predefined categories to text documents, used in spam filtering, sentiment analysis, and topic labelling.
An NLP task that produces a shorter version of a document while preserving its key information, either extractive or abstractive.
The number of tokens or requests an AI inference system can process per unit of time, a key metric for scaling production deployments.
The ability of an LLM to call external functions, APIs, or services during generation, enabling real-time data retrieval and action execution.
A sampling method that restricts token selection to the smallest set whose cumulative probability exceeds P, balancing diversity and coherence.
The dataset used to optimise a model's parameters during the learning phase, directly shaping its capabilities and potential biases.
