Techniques & Methods

The ability of a model to correctly perform tasks it has not explicitly been trained to do, demonstrating generalization.

A technique in NLP where words are represented as vectors in a high-dimensional space, capturing semantic similarity.

The encoding of words or phrases as numerical vectors, enabling mathematical operations and comparisons by AI models.

Different expressions or phrasings that convey the same intent or meaning, important in understanding natural language variability.

The process of evaluating a model's performance with a separate portion of the data not used in training, to gauge its accuracy.

A technique in generative AI where multiple outputs are generated and the best one is selected based on certain criteria.

Leveraging knowledge gained while solving one problem to solve a different but related problem in machine learning.

The process of teaching a machine learning model to make predictions or decisions, typically by exposing it to a large dataset.

A statistical model to discover abstract topics within a collection of documents, aiding in content organization and discovery.

The task of assigning predefined categories to text, used in applications like spam detection and sentiment analysis.

Internal cues or instructions that guide the behavior of an AI model, influencing how it processes and responds to input.

The process of refining a model's performance on specific tasks by training it further with labeled data.

The process where AI models produce a sequence of items, such as words in text generation, based on learned patterns.

The measure of how much two pieces of text are related in terms of meaning, used in various NLP tasks.

The process of adding semantic metadata to content, making it easier for AI to understand and process information.

A mechanism that allows models to weigh the importance of different parts of the input data relative to each other.

Observations that as AI models increase in size, their performance improves according to predictable patterns.

Combining retrieval of relevant information with generative models to produce informed responses.

An evaluation of how well an AI system's responses meet the criteria of relevance, coherence, and accuracy.

Training approach where models are refined based on feedback from human evaluators.

Techniques used to prevent overfitting by penalizing complex models during the training process.

A request for information or action made to a database, search engine, or AI model.

A reinforcement learning algorithm that balances exploration and exploitation in policy learning.

A technique used to influence or manipulate the behavior of AI systems through specially crafted inputs.

The art of crafting prompts to effectively communicate with and elicit desired responses from AI models.

A text input given to an AI model, designed to elicit a specific type of response or output.

The initial training phase where a model learns from a large, general dataset before task-specific training.

The process of marking up a word in a text as corresponding to a particular part of speech.

A technique to discourage repetitive or overly similar responses in generative AI models.

A model training approach where the model updates continuously as new data arrives.

The ability of a model to learn information from a single example or a few examples.

Learning techniques where the model learns from one or a few examples, respectively.

Learning optimal actions from a fixed dataset without further interaction with the environment.

The process of identifying and classifying key information (entities) in text into predefined categories.

Training an AI model on multiple tasks simultaneously, leveraging commonalities across tasks.

A training technique where some words in the input are hidden, and the model predicts them.

A mathematical framework for modeling decision-making in situations with random outcomes.

A technique for fine-tuning large models in a memory and computationally efficient manner.

The process of adding metadata regarding linguistic information to text, aiding in its analysis.

The method by which AI systems model, store, and retrieve knowledge to solve complex tasks.

The probability of two events happening at the same time in a probabilistic model.

The process of automatically extracting structured information from unstructured text data.

The phase where a trained model is used to make predictions or decisions based on new, unseen data.

Problem-solving approaches that use practical methods or various shortcuts to produce solutions.

When AI generates information that is not grounded in reality, often due to training data issues.

Optimization algorithms that make the locally optimal choice at each step to find a global optimum.

The process of producing new content, such as text or images, based on learned patterns and data.

A logical reasoning method that starts with known facts and applies rules to reach new conclusions.

The process of adjusting a pre-trained model to perform well on a specific task or dataset.

The capability to precisely adjust the output or behavior of an AI model based on specific criteria.

The ability of a model to learn and generalize from a very small number of examples.

Identifying and isolating useful information from data to improve model training and performance.

Creating summaries by extracting key sentences or fragments directly from the source text.

Quantitative measures used to assess the performance and effectiveness of AI models.

Identifying and classifying named entities in text into predefined categories.

The process of labeling text with information about entities, enhancing data structure.

A method where AI model training is spread across multiple computers or servers.

Analyzing the grammatical structure of a sentence to understand relationships between words.

Guidelines that dictate how a language model translates its internal representations to output.

The practice of examining large databases to generate new information and find hidden patterns.

A technique for increasing the amount of training data by adding slightly modified copies.

The task in NLP of determining which words refer to the same entity in a text.

The output produced by AI in response to a given input or prompt, completing the thought process.

A prompting strategy that encourages AI to break down complex problems into manageable steps.

A search algorithm that efficiently finds the most likely sequences of outcomes in models.

A strategy for balancing the exploration of new choices and the exploitation of known rewards.

A reasoning method that starts with the end goal and works backward to determine the solution path.

A method used in training artificial neural networks, adjusting weights based on error rates.

A statistical model that predicts future behavior based on past outcomes in time series data.

In AI, a technique that helps models focus on relevant parts of the input data, improving relevance.

A mechanism in AI that allows models to weigh the importance of different pieces of information.

The process of ensuring AI behaviors and outputs adhere to human ethical standards and intentions.

Improves AI robustness by training with deliberately challenging inputs to enhance model accuracy.