Comprehensive reference for all key terms and mathematical formulas
Neural network architecture using self-attention mechanisms for parallel sequence processing.
Process of splitting text into smaller units (tokens) for model processing.
Component that allows models to weigh the importance of different input parts.
AI model trained on vast text data with billions of parameters.
Process of further training a pre-trained model on specific domain data.
Numerical representation of text tokens in vector space.
Technique of designing inputs to guide AI model outputs.
Model performing tasks without specific training examples.
Model learning from a small number of examples provided in the prompt.
Parameter controlling randomness in model output (0=deterministic, 1=creative).
Assigning a specific persona or role to the AI model.
Decomposing text into isolated style features (tone, length, etc.).
Prompting technique encouraging step-by-step reasoning.
Using prompts to generate or optimize other prompts.
Survey question asking about two things at once.
Psychometric scale for measuring attitudes and opinions.
Combining reasoning and acting in an iterative loop.
Autonomous system that reasons, plans, and acts using tools.
Framework for building applications with LLMs and tools.
API feature enabling structured data extraction with defined schemas.
Capability of AI models to interact with external systems and APIs.
Component allowing agents to retain and recall past interactions.
Generative model that creates images by reversing a noise-adding process.
Process of progressively removing noise to generate coherent outputs.
Compressed representation where diffusion process occurs.
Extension allowing structural control over diffusion model outputs.
Instructions specifying what the model should NOT include.
Technique for filling in missing or selected parts of an image.
Technique for extending an image beyond its original borders.
Image-to-image generation using reference images as starting point.
Practice of optimizing content for search engine visibility.
Frequency of target keywords in content relative to total word count.
Metric measuring how easy text is to read (e.g., Flesch score).
Comparing two versions to determine which performs better.
HTML element defining the title shown in search results.
Summary text shown below title in search engine results.
Link from another website pointing to your content.
Experience, Expertise, Authoritativeness, Trustworthiness (Google quality guidelines).
Set of protocols for building software applications.
AI generating plausible but factually incorrect information.
Adjustable value in a model that is learned during training.
Process of using a trained model to make predictions.
Maximum number of tokens a model can process in one request.
Measures similarity between two vectors.
Core transformer computation.
Frequency of itemset.
Conditional probability.
Strength of association.
Harmonic mean.
Accuracy of positive predictions.
Coverage of actual positives.
Overall correctness.
Controls randomness.
Distance between two points.
Heuristic for cluster count selection.
Measure of node impurity.
Term importance measure.
Classification loss.