Paradigm on English AI Terms Dictionary

Instruction Following

Sat, 18 Jul 2026 10:02:57 +0000

Definition

Instruction following refers to the ability of large language models and other AI systems to understand nuanced human directives and adhere to explicit constraints within a prompt. This paradigm shifts interaction from open-ended generation to task-specific execution, ensuring outputs align precisely with user intent, format requirements, and logical boundaries. It is foundational for reliable AI integration in professional workflows where precision and compliance are critical.

Summary

The capability of an AI model to accurately interpret and execute specific human commands or constraints.

Embodied agent

Sat, 18 Jul 2026 09:56:39 +0000

Definition

Unlike disembodied AI that processes abstract data, embodied agents learn and act within a physical context, relying on sensory inputs and motor outputs. This paradigm is central to robotics and autonomous systems, where intelligence emerges from the interaction between the agent’s body, its control mechanisms, and the surrounding world. It emphasizes that cognition is deeply rooted in physical experience.

Summary

An embodied agent is an AI system that interacts with its environment through a physical body or sensorimotor apparatus.

Case-based reasoning

Sat, 18 Jul 2026 09:48:49 +0000

Definition

CBR operates on the principle that similar problems have similar solutions. The process involves retrieving the most similar historical case from a knowledge base, adapting its solution to fit the current context, and retaining the new experience for future use. This paradigm is particularly useful in domains where explicit rules are difficult to define, such as legal reasoning, medical diagnosis, and customer service automation, leveraging experiential knowledge rather than purely symbolic logic.

learning-based

Sat, 18 Jul 2026 09:38:47 +0000

Definition

Learning-based approaches rely on statistical algorithms to identify patterns and make decisions based on data exposure, contrasting with rule-based systems. This category encompasses supervised, unsupervised, and reinforcement learning techniques. By optimizing objective functions through iterative updates, these systems adapt to new information. This paradigm is central to modern AI, allowing for flexibility and automation in tasks ranging from image recognition to strategic game playing.

Summary

Indicates methods where algorithms improve performance through experience rather than explicit programming rules.