18 terms
Na total amount of text or other information wey AI model fit consider for one time, including the conversation wey matter.
The amount of text or other input an AI model can consider at one time, including relevant conversation history.
Na when you divide difficult request into smaller tasks wey AI fit handle more easily.
Breaking a complex request into smaller, easier-to-handle tasks for an AI system.
Na prompt wey get small number of examples to guide model make e do task or follow answer format.
A prompt that includes a small number of examples to guide a model toward a task or answer format.
Na when you give AI useful, trusted information make the answer connect to wetin you provide.
Providing relevant, trusted information in the input so an AI answer is tied to that information.
Na when AI talk something with confidence, but the information no true, no get evidence, or na im make am up.
An AI output that sounds confident but contains false, unsupported, or invented information.
Na words or instruction wey tell image model wetin e no suppose put inside the picture.
Words or instructions describing what an image-generation model should avoid including in its output.
Na prompt wey get one example to show model the kind work or answer format wey you expect.
A prompt that includes one example to show a model the expected task or answer format.
Na instruction, question, or information wey you give AI to guide the answer wey e go give you.
The instruction, question, or information a person gives an AI system to guide its response.
Na when you divide big work into prompts, then use answer from one step as input for the next step.
Breaking a larger task into prompts where the output of one step becomes input for the next.
Na how person dey write and adjust instruction so AI go get better chance to give useful answer.
The practice of writing and refining instructions so an AI system is more likely to produce a useful result.
Na attempt to make AI ignore or change im instruction by putting conflicting command inside user input or information wey e retrieve.
An attempt to override an AI system’s instructions by placing conflicting instructions in user input or retrieved content.
Na prompt wey you fit reuse, with fixed instruction and blank spaces wey you go fill with different information.
A reusable prompt with fixed instructions and placeholders that can be filled with different information.
Na when you give model role or point of view, like “act as patient teacher,” to shape the answer.
Giving a model a role or point of view, such as “act as a patient tutor,” to shape its response.
Na AI answer wey follow one format wey dem set, like JSON, table, or list of named fields.
An AI response constrained to a format such as JSON, a table, or a list of named fields.
Na important instruction wey set how model go behave, the role wey e get, or the limits wey e must follow for conversation or app.
High-priority instructions that set a model’s overall behavior, role, or constraints for a conversation or application.
Na model setting wey fit affect how varied or predictable the generated text go be; how e work depend on the model.
A model setting that influences how varied or predictable generated text is; its effect depends on the model.
Na generation setting wey limit the possible next tokens to a group based on probability before model choose one.
A generation setting that limits the next-token choices to a probability-based group before the model selects one.
Na prompt wey ask model to do work without showing am examples of the kind answer wey you want.
A prompt that asks a model to do a task without showing it examples of the desired answer.