Start learning

ML and AI Glossary

51 terms, defined the way they are used in engineering decisions rather than textbooks.

Read as one page

A

  • A/B Test: An experiment that randomly splits users into two groups to compare two interventions. The control group gets the current behavior; the…
  • A2A (Agent-to-Agent Protocol): A protocol for one agent to discover another (via a published "agent card"), delegate a task to it, and receive results. Complements MCP…
  • Agent Harness: The deterministic code around a model that turns it into an agent: the tool list, the permission and budget checks, the verification…
  • AUC-PR (Area Under the Precision-Recall Curve): Measures classifier performance specifically on the positive class. More informative than AUC-ROC when positive examples are rare. A model…
  • AUC-ROC (Area Under the Receiver Operating Characteristic Curve): Measures how well a binary classifier separates positive and negative classes across all thresholds. A value of 1.0 is perfect; 0.5 is…

B

  • Batch Normalization: A technique that normalizes layer activations during training, reducing internal covariate shift. Applied after a linear layer and before…

C

  • Checkpointing: Saving an agent's full run state (current step, tool results so far, pending approvals) after every step so the run can resume from the…
  • Cold Start: The problem of making predictions for new users or new items with no historical interaction data. Common in recommendation systems.…
  • Compaction: Summarizing or dropping earlier context when the window fills. Always lossy; the engineering is in deciding what to keep, keeping raw…
  • Computer-Use Agent: An agent that operates a graphical interface the way a person does: it looks at a screenshot, decides where to click or type, and acts.…
  • Confusion Matrix: A 2×2 table for binary classification: True Positives, True Negatives, False Positives, False Negatives. From these four numbers, you can…
  • Context Engineering: The discipline of designing everything the model sees on a given turn: system prompt, tool definitions, retrieved documents, memory, and…
  • Cross-Entropy Loss: The standard loss function for classification. For binary classification: -y·log(p) + (1-y)·log(1-p). For multi-class: -Σ y_i·log(p_i).…

D

  • Data Leakage: When information from the future or from the target variable leaks into the training features, inflating offline metrics. The model…
  • DPO (Direct Preference Optimization): A post-training method that trains a model directly on pairs of (preferred, rejected) responses without a separate reward model. Simpler…
  • Dropout: A regularization technique where a random fraction of neurons are set to zero during training. This prevents neurons from co-adapting too…
  • Durable Execution: Running a multi-step workflow so that it survives process crashes, restarts, and long waits (for a human, for an external system) without…

E

  • Early Stopping: Stopping training when validation loss stops improving, instead of training for a fixed number of epochs. Prevents overfitting without…
  • Embedding: A dense vector representation of a discrete entity (word, user, product, category). Learned from data such that similar entities are close…
  • Excessive Agency: The failure where an agent has more permissions or tools than its task requires, so that a mistake or an injection can cause…

F

  • F1 Score: The harmonic mean of precision and recall: 2 · precision · recall / (precision + recall). Use when both false positives and false…
  • Feature Importance: A measure of how much each feature contributes to model predictions. For tree models (Random Forest, XGBoost): typically measures the…
  • Fine-Tuning: Adapting a pretrained model to a new task by continuing training on task-specific data. The pretrained weights serve as initialization.…

G

  • Gradient Vanishing / Exploding: Problems in training deep networks where gradients become extremely small (vanishing) or large (exploding) during backpropagation, making…
  • GRPO (Group Relative Policy Optimization): A reinforcement learning algorithm for post-training that samples a group of responses per prompt and uses each response's reward relative…
  • Guardian Pattern: Placing a second, narrowly scoped model or rule set between the agent and its tools (or between untrusted content and the agent) to…

H

  • Human-in-the-Loop (HITL): Designing the workflow so that a person approves specific actions (usually irreversible ones) before they execute. In a durable agent, the…
  • Hyperparameter: A parameter that controls the learning process and is set before training, not learned from data. Examples: learning rate, number of…

L

  • LLM-as-Judge: Using a model to grade another model's output against a rubric. Cheap and scalable; must itself be validated against human labels on a…

M

  • MCP (Model Context Protocol): The open standard for connecting a model to tools, data, and prompts. An MCP server exposes tools (callable functions), resources…
  • Memory Poisoning: An attack or accident where bad content enters an agent's long-term memory and then influences every future run. The reason long-term…
  • Model Routing: Sending each request to the cheapest model that can handle it, based on rules, a classifier, or a first attempt with a small model plus a…

N

O

  • Overfitting: When a model learns the training data too well, including its noise, and fails to generalize to new examples. Symptoms: very low training…

P

  • Precision: Of all the examples predicted positive, what fraction are actually positive? TP / (TP + FP). High precision means few false alarms.…
  • Prompt Caching: Reusing the computed state (the KV cache) for a prompt prefix across requests. Providers bill cached tokens at a discount; serving…
  • Prompt Injection: Content the model reads (a web page, an email, a document, a tool result) that contains instructions the model then follows. Structurally…

R

  • Reasoning Model: A model post-trained (usually with RLVR) to spend tokens on intermediate reasoning before answering. Better on verifiable, multi-step…
  • Recall (Sensitivity, True Positive Rate): Of all the actual positive examples, what fraction did the model find? TP / (TP + FN). High recall means few misses. Critical when false…
  • Regularization: Techniques that reduce model complexity to improve generalization. L2 (Ridge): penalizes large weights, drives them toward zero. L1…
  • RLVR (Reinforcement Learning with Verifiable Rewards): Post-training where the reward is computed by a checker (unit tests pass, the numeric answer matches, the output parses) instead of a…

S

  • Sandbox: An isolated execution environment (container, VM, restricted interpreter) where an agent's code or tool calls run so that a mistake or an…
  • SLM (Small Language Model): A model in roughly the 1B to 8B parameter range that runs on a laptop, a phone, or a single small GPU. Competitive with older large models…
  • Speculative Decoding: A serving technique where a small draft model proposes several tokens and the large model verifies them in one pass, accepting the…

T

  • Test-Time Compute: Spending more inference compute on a single request to get a better answer: longer reasoning, sampling several answers and voting, or…
  • Tokenization: Converting text into tokens - the atomic units a language model processes. Common approaches: word tokenization (splits on spaces, loses…
  • Training-Serving Skew: When the feature values used during training differ from those used at inference time. Causes: different code paths for training vs.…
  • Trajectory: The full sequence of an agent's reasoning, tool calls, tool results, and intermediate outputs across a run. Agent evaluation grades the…

U

  • Underfitting: When a model is too simple to capture the patterns in the data. Symptoms: high training loss and high validation loss. Solutions: more…

V

  • Validation Set: A held-out subset of data used to tune hyperparameters and make modeling decisions during development. Distinct from the test set, which…
  • Verifiable Reward: A reward you can compute mechanically: tests pass, the answer matches, the schema validates. The precondition for RLVR and the cheapest…