API

Application Programming Interface - structured interface for software communication.

Week 1

Agent

An autonomous system that perceives its environment, reasons about it, and takes actions to achieve goals.

Week 1

AgentBench

Benchmark for evaluating LLMs as agents across multiple environments.

Week 10

Alignment

Training AI systems to follow human values and intentions.

Week 12

AutoGPT

Goal-directed autonomous agent that chains LLM calls for open-ended tasks.

Week 1

AutoGen

Microsoft's framework for building multi-agent conversational systems.

Week 6

BabyAGI

Task-driven autonomous agent using task prioritization and queuing.

Week 1

CRAG

Corrective RAG - system that evaluates and corrects poor retrievals before generation.

Week 7

Chain-of-Thought (CoT)

Prompting technique that elicits step-by-step reasoning from language models.

Week 2

Chain-of-Verification

Technique to reduce hallucinations through independent verification of claims.

Week 9

Claim Decomposition

Breaking down text into atomic, independently verifiable claims.

Week 9

Claude

Anthropic's large language model with Constitutional AI safety features.

Week 1

Code Agent

Agent specialized for code generation, testing, and debugging.

Week 11

Constitutional AI

Approach to AI safety using principles to guide self-improvement.

Week 12

Context Window

Maximum number of tokens an LLM can process in a single interaction.

Week 1

CrewAI

Framework for building role-playing autonomous AI agents that collaborate.

Week 6

Deliberative Architecture

Agent architecture that plans before executing (model-based reasoning).

Week 1

Devin

Cognition's autonomous software engineering agent.

Week 1

Domain Agent

Agent designed for a specific domain with specialized tools and constraints.

Week 11

Embodied Agent

Agent that interacts with a physical or simulated environment through sensors and actuators.

Week 12

Emergent Behavior

Complex behaviors arising from simple agent interactions.

Week 12

Entity Extraction

Process of identifying and extracting named entities from text.

Week 8

Episodic Memory

Memory system storing past experiences and reflections for future reference.

Week 4

FActScore

Metric for evaluating factual precision in long-form text generation.

Week 9

Function Calling

API feature allowing LLMs to generate structured calls to predefined functions.

Week 3

GPT-4

OpenAI's large language model powering ChatGPT and API.

Week 1

Gemini

Google's multimodal large language model.

Week 1

Generative Agent

Agent that simulates human behavior in interactive environments.

Week 12

GitHub Copilot

AI-powered code completion and generation tool by GitHub/Microsoft.

Week 1

GraphRAG

RAG approach using knowledge graphs for structured, multi-hop retrieval.

Week 8

Hallucination

When an LLM generates plausible but factually incorrect information.

Week 9

Hierarchical Architecture

Architecture where manager agents delegate to worker agents.

Week 1

Hybrid Architecture

Agent architecture combining reactive and deliberative layers.

Week 1

Jailbreak

Technique to bypass LLM safety restrictions through adversarial prompts.

Week 12

Knowledge Graph

Graph-based representation of entities and their relationships.

Week 8

LATS

Language Agent Tree Search - combines reasoning with Monte Carlo Tree Search for planning.

Week 4

LLM

Large Language Model - neural network trained on text to generate and understand language.

Week 1

LLM-as-Judge

Using an LLM to evaluate the quality of outputs from another model or agent.

Week 10

LangChain

Python framework for building LLM applications using chains of components.

Week 6

LangGraph

LangChain's framework for building stateful, multi-actor applications using graphs.

Week 6

Long-term Memory

Persistent storage across sessions using vector DBs or knowledge graphs.

Week 4

Memory

System for storing and retrieving information across agent interactions.

Week 1

Message Passing

Communication pattern where agents exchange structured messages.

Week 5

Model Context Protocol (MCP)

Anthropic's open protocol for connecting LLMs to external tools and data sources.

Week 3

Multi-Agent System

System with multiple agents that coordinate to solve complex tasks.

Week 5

Observation

Feedback from the environment after an agent takes an action.

Week 1

Orchestration

Coordination pattern where a central agent manages workflow across multiple agents.

Week 5

Pass@k

Metric measuring success rate when k attempts are allowed.

Week 10

Planning

The process of decomposing complex tasks into manageable steps before execution.

Week 4

RAG

Retrieval-Augmented Generation - combining retrieval with generation for grounded responses.

Week 7

RAPTOR

Recursive Abstractive Processing for Tree-Organized Retrieval - hierarchical RAG approach.

Week 7

ReAct

Reasoning and Acting paradigm that interleaves thought, action, and observation steps.

Week 1

Reactive Architecture

Agent architecture using direct stimulus-response without planning.

Week 1

Reflexion

Framework where agents learn from verbal self-reflection on past failures.

Week 4

SWE-bench

Benchmark for evaluating code agents on real-world GitHub issues.

Week 10

Self-Consistency

Sampling multiple reasoning paths and selecting the most common answer through majority voting.

Week 2

Self-RAG

RAG system that learns when to retrieve and how to critique its own outputs.

Week 7

Semantic Kernel

Microsoft's SDK for integrating LLMs with conventional programming.

Week 6

Short-term Memory

In-context memory stored within the current conversation window.

Week 4

Single-Agent System

Architecture where one LLM handles all tasks.

Week 1

Task Decomposition

Breaking complex tasks into smaller, manageable subtasks.

Week 4

Token

Basic unit of text processing in LLMs, roughly equivalent to 3/4 of a word.

Week 1

Tool Use

The ability of an agent to invoke external tools or APIs to accomplish tasks.

Week 3

Trajectory

The complete sequence of states, actions, and observations an agent takes during task execution.

Week 1

Tree-of-Thoughts (ToT)

Framework that explores multiple reasoning paths and evaluates them to find optimal solutions.

Week 2

Vector Store

Embedding database for semantic similarity search used in RAG systems.

Week 7

Voyager

Open-ended embodied agent for Minecraft exploration (NVIDIA/Microsoft).

Week 1

World Model

Internal representation of environment dynamics used for planning and simulation.

Week 4

Glossary

API

Agent

AgentBench

Alignment

AutoGPT

AutoGen

BabyAGI

CRAG

Chain-of-Thought (CoT)

Chain-of-Verification

Claim Decomposition

Claude

Code Agent

Constitutional AI

Context Window

CrewAI

Deliberative Architecture

Devin

Domain Agent

Embodied Agent

Emergent Behavior

Entity Extraction

Episodic Memory

FActScore

Function Calling

GPT-4

Gemini

Generative Agent

GitHub Copilot

GraphRAG

Hallucination

Hierarchical Architecture

Hybrid Architecture

Jailbreak

Knowledge Graph

LATS

LLM

LLM-as-Judge

LangChain

LangGraph

Long-term Memory

Memory

Message Passing

Model Context Protocol (MCP)

Multi-Agent System

Observation

Orchestration

Pass@k

Planning

RAG

RAPTOR

ReAct

Reactive Architecture

Reflexion

SWE-bench

Self-Consistency

Self-RAG

Semantic Kernel

Short-term Memory

Single-Agent System

Task Decomposition

Token

Tool Use

Trajectory

Tree-of-Thoughts (ToT)

Vector Store

Voyager

World Model