Skip to main content
Product
Docs
Pricing
Contact
Resources
Docs
Dashboard
AI & LLM Glossary
Clear definitions and explanations of AI, LLM, and machine learning concepts.
All
A
B
C
D
E
F
G
H
I
J
K
L
M
N
O
P
Q
R
S
T
U
V
Z
A
Adversarial Attacks
Adversarial attacks manipulate AI inputs to cause incorrect predictions. Learn attack types, real-world examples, and defense strategies for ML models and LLMs.
Agentic AI
Agentic AI describes AI systems that autonomously plan, execute multi-step tasks, and use tools to achieve goals. Discover how AI agents work in practice.
Agents
AI agents are autonomous systems that perceive, reason, plan, and act to achieve goals. Learn how AI agents work, their architectures, and real-world applications.
Alignment
AI alignment ensures AI systems behave according to human values and intentions. Learn alignment techniques, challenges, and why it matters for safe AI.
Alignment
Learn what alignment means in AI and LLMs, why it matters for safe AI systems, and how techniques like RLHF help align models with human values.
Agent
Learn what an AI agent is, how agents use LLMs to reason and take actions autonomously, and why agentic systems are transforming AI applications.
Autoregressive Model
Learn what autoregressive models are, how they generate text token by token, and why this approach powers modern LLMs like GPT, Claude, and Llama.
Attention Mechanism
Understand attention mechanisms in AI, how they enable transformers to process context, and why they are foundational to modern LLMs like GPT and Claude.
B
Batching
Learn what batching means in AI inference, how it improves LLM throughput and reduces costs, and best practices for implementing batch processing.
Benchmarking
Learn what benchmarking means in AI, how LLM benchmarks like MMLU and HumanEval measure model capabilities, and how to evaluate models for your use case.
Bias Detection
Learn what bias detection means in AI, how to identify and mitigate bias in LLM outputs, and why it is essential for building fair and responsible AI systems.
C
Compliance
AI compliance is the practice of ensuring AI systems meet legal, regulatory, and ethical standards. Learn about frameworks, requirements, and best practices.
Caching
Learn what caching means for LLMs, how prompt caching and KV-caching reduce latency and cost, and best practices for implementing caching in AI applications.
Catastrophic Forgetting
Catastrophic forgetting occurs when neural networks lose previously learned knowledge while training on new tasks. Learn causes, solutions, and mitigation strategies.
Chain of Thought
Learn what chain of thought prompting is, how it improves LLM reasoning abilities, and practical techniques for getting better results from AI models.
Chunking
Learn what chunking means in AI, how to split documents for RAG and vector search, and best practices for chunk sizes and strategies in LLM applications.
Context Window
Learn what a context window is in LLMs, how context length affects model capabilities, and strategies for working within or extending context limits.
Caching
Learn what LLM caching is, how it reduces latency and cost by reusing previous model responses, and best practices for implementing caching in LLM applications.
D
Data Poisoning
Learn what data poisoning is in AI, how attackers corrupt training data to manipulate model behavior, and how to defend against it.
Distillation
Learn what knowledge distillation is in AI, how smaller models learn from larger ones, and why it matters for efficient LLM deployment.
E
Embeddings
Learn what embeddings are in AI, how text and data are converted to numerical vectors, and why they power search, RAG, and similarity tasks.
Evaluation Metrics
Learn what evaluation metrics are in AI, how they measure model performance, and which metrics matter most for LLM applications.
Explainability
Learn what explainability means in AI, how it helps users understand model decisions, and why it is essential for trustworthy AI systems.
Explainable AI
Explainable AI (XAI) makes AI decisions transparent and understandable to humans. Learn about XAI techniques, methods, and why interpretability matters for trust.
F
Few-shot Learning
Learn what few-shot learning is in AI, how models learn from just a few examples, and how it differs from zero-shot and fine-tuning approaches.
Fine-tuning
Learn what fine-tuning is in AI, how it adapts pre-trained models for specific tasks, and when to use it vs. prompting or RAG.
Function Calling
Learn what function calling is in AI, how LLMs interact with external tools and APIs, and why it enables powerful AI agent capabilities.
G
Guardrails
Learn what guardrails are in AI, how they enforce safe and compliant model behavior, and why they are essential for production LLM deployments.
Gateway
Learn what an AI gateway is, how it acts as a unified proxy for LLM API traffic, and why it is critical for managing cost, reliability, and security at scale.
Governance
AI governance is the framework of policies, processes, and controls for managing AI systems responsibly. Learn about governance structures, principles, and implementation.
Guardrails
Learn what AI guardrails are, how they enforce safety and compliance in LLM applications, and why they are essential for responsible AI deployment.
Gateway
Learn what an LLM gateway is, how it provides a unified API layer for managing multiple LLM providers, and why it simplifies AI infrastructure at scale.
Governance
Learn what AI governance means, how it establishes policies for responsible AI use, and why structured oversight is essential for safe LLM deployments.
Grounding
Learn what grounding is in AI, how it connects LLM outputs to verified sources, and why it reduces hallucinations in production systems.
H
Hallucination
AI hallucination occurs when a language model generates confident-sounding but factually incorrect or fabricated information not grounded in its training data or context.
Hallucination Detection
Learn what hallucination detection is, how it identifies false or fabricated content in LLM outputs, and why it is critical for building trustworthy AI applications.
Human Feedback
Learn what human feedback means in AI, how it improves LLM outputs through ratings and corrections, and why it is critical for aligning models with human values.
I
In-context Learning
Learn what in-context learning means, how LLMs learn from examples provided in the prompt without retraining, and when to use it versus fine-tuning.
Inference Latency
Learn what inference latency means in AI, how it affects LLM response times, and practical strategies to reduce latency in production AI applications.
J
JSON Mode
Learn what JSON mode is in LLMs, how it ensures models output valid JSON, and why structured output is essential for building reliable AI applications.
K
Knowledge Graph
Learn what a knowledge graph is, how it structures information as entities and relationships, and how it enhances LLM accuracy through grounded knowledge retrieval.
L
LoRA
Learn what LoRA (Low-Rank Adaptation) is, how it enables efficient LLM fine-tuning with minimal resources, and why it has become the standard for model customization.
M
Model Card
A model card is a standardized document that describes an AI model's capabilities, limitations, training data, and intended use. Learn why model cards matter.
Mixture of Experts
Learn what Mixture of Experts (MoE) is, how it scales LLMs efficiently by activating only a subset of parameters per input, and its impact on AI performance.
Model Cards
Model cards are structured documentation for AI models covering performance, limitations, and intended use. Learn how they promote transparency and accountability.
Model Collapse
Model collapse occurs when AI models trained on synthetic or AI-generated data progressively degrade. Learn the causes, stages, and prevention strategies.
Model Drift
Model drift occurs when an AI model's performance degrades over time due to changes in data patterns. Learn causes, types, and how to detect it.
Model Evaluation
Model evaluation is the systematic process of measuring an LLM's output quality, accuracy, and safety using automated metrics, human review, and benchmark testing.
Model Serving
Model serving is the process of deploying trained ML models to production so they can handle real-time predictions. Learn about serving infrastructure and best practices.
Multimodal AI
Multimodal AI refers to systems that process and generate multiple data types like text, images, audio, and video. Learn how it works and why it matters.
Multimodal
Learn what multimodal AI means, how models process text, images, audio, and video together, and why multimodal capabilities are transforming AI applications.
N
Neural Architecture Search
Neural Architecture Search (NAS) automates the design of neural network architectures. Learn how NAS works, its methods, and applications in AI.
O
Orchestration
AI Orchestration is the practice of coordinating multiple AI models, tools, and data sources into unified workflows. Learn key patterns and best practices.
Observability
LLM observability is the practice of monitoring, tracing, and analyzing the behavior and performance of large language model applications in production.
Orchestration
Orchestration in AI coordinates multiple models, tools, and workflows into cohesive systems. Learn how orchestration enables complex AI applications.
P
Prompt Chaining
Prompt chaining connects multiple LLM calls in sequence, where each output feeds into the next. Learn how it enables complex AI workflows.
Prompt Engineering
Prompt engineering is the practice of crafting effective inputs for LLMs to produce desired outputs. Learn techniques, best practices, and real-world examples.
Prompt Injection
Prompt injection is an attack technique where malicious input manipulates a large language model into ignoring its instructions or producing unintended outputs.
Prompt Optimization
Prompt Optimization is the systematic process of refining LLM prompts to improve output quality, reduce costs, and increase reliability. Learn proven techniques.
Prompt Template
Learn what prompt templates are, how they standardize LLM interactions with reusable parameterized prompts, and best practices for template design.
Q
Quantization
Quantization reduces AI model size by using lower-precision numbers for weights and computations. Learn about quantization methods and their impact on LLMs.
R
RAG
RAG (Retrieval-Augmented Generation) enhances LLM responses by retrieving relevant documents before generating answers. Learn how RAG works and why it matters.
Re-ranking
Re-ranking reorders initial search results using a more powerful model to surface the most relevant documents for a query.
Red Teaming
Red teaming is the practice of deliberately probing AI systems for vulnerabilities, biases, and failure modes to improve safety and robustness.
Responsible AI
Responsible AI is the practice of developing and deploying AI systems that are fair, transparent, accountable, and aligned with ethical standards.
Retrieval-Augmented Generation (RAG)
Learn what Retrieval-Augmented Generation (RAG) is, how it works by combining LLM generation with external knowledge retrieval, and why it reduces hallucinations.
RLHF
RLHF (Reinforcement Learning from Human Feedback) aligns AI models with human preferences through reward modeling and policy optimization. Learn how it works.
S
Safety
AI Safety is the field dedicated to ensuring AI systems operate reliably, align with human values, and avoid harmful outcomes. Learn key concepts and methods.
Safety
AI safety is the field dedicated to ensuring AI systems operate reliably and do not cause unintended harm to users or society.
Semantic Search
Semantic search finds results based on the meaning of a query rather than exact keyword matches, using embeddings and vector similarity.
Streaming
Streaming delivers LLM responses token-by-token in real time rather than waiting for the complete response, improving perceived latency.
Structured Output
Structured output constrains LLM responses to follow a specific format like JSON, ensuring reliable parsing and integration with downstream systems.
Summarization
Summarization uses AI to condense long texts into shorter versions that capture the key information, saving time and improving comprehension.
System Prompt
Learn what a system prompt is, how it steers LLM behavior by defining roles and constraints, and best practices for writing effective system prompts.
T
Transformer
The transformer is the neural network architecture behind modern LLMs, using self-attention to process and generate text with remarkable effectiveness.
Temperature (LLM)
Learn what temperature means in LLMs, how it controls output randomness and creativity, and best practices for setting temperature in different AI applications.
Temperature
Temperature is a parameter that controls the randomness of LLM outputs, with lower values producing focused responses and higher values increasing creativity.
Token Cost
Token cost is the price charged by LLM providers per token processed, covering both input (prompt) and output (completion) tokens in API-based AI applications.
Token Limit
Learn what token limits are in LLMs, how they constrain input and output length, and strategies for working within context window boundaries effectively.
Tokenization
Tokenization splits text into smaller units called tokens that LLMs can process, directly affecting model performance, cost, and context limits.
Tool Use
Tool use enables LLMs to interact with external tools, APIs, and systems, extending their capabilities beyond text generation.
Transfer Learning
Transfer learning reuses knowledge from a pre-trained model to solve new tasks faster with less data. Learn how it works, key examples, and LLM applications.
Transformer Architecture
Learn what the Transformer architecture is, how self-attention mechanisms work, and why Transformers are the foundation of modern LLMs like GPT and Claude.
U
Uncertainty Estimation
Uncertainty estimation quantifies how confident an AI model is in its predictions, helping identify unreliable outputs and improve decision-making.
V
Vector Database
A vector database stores and efficiently searches high-dimensional vectors, enabling fast similarity search for AI applications like RAG and recommendations.
Z
Zero-shot Learning
Zero-shot learning enables AI models to perform tasks they were never explicitly trained on by leveraging general knowledge and natural language instructions.