AI Glossary

The computational component of an AI system that uses learned parameters, encoded knowledge, or both to infer outputs from inputs.

The operational arrangement that uses one or more AI models with data, software, infrastructure, interfaces, controls, and people to produce and act on outputs.

A specified procedure that transforms inputs into outputs or solves a defined class of problems through a finite set of steps and stated stopping behavior.

A contested term for AI with broad, transferable competence across many cognitive tasks rather than capability limited to one task or domain.

The field concerned with machine-based systems that infer how to produce predictions, content, recommendations, decisions, or actions in pursuit of stated or implicit objectives.

A learned mechanism that scores relationships among elements and uses those scores to combine information into a new representation.

Assigning one or more discrete categories to an input, often by converting model scores or probabilities into labels through a decision rule.

Grouping examples without predefined class labels according to a chosen representation, similarity measure, and clustering objective.

The field of building computational systems that derive measurements, representations, predictions, or actions from images, video, and other visual signals.

The token capacity a model can process within one inference request, shared among the input context and, for many systems, the tokens generated in response.

An assembled collection of examples or records, with an intended structure and use, for training, tuning, evaluating, or operating an AI system.

Machine learning based on neural networks with many successive layers that learn increasingly useful representations from data.

A numeric vector produced by a representation model so useful relationships among items can be compared with geometric operations.

A bounded AI system that applies an explicit knowledge base and inference procedure to reproduce part of a specialist's decision process.

An input variable or derived representation that a machine-learning model actually uses to produce a prediction.

A model trained on broad data, generally through self-supervision at scale, that can be adapted to a wide range of downstream tasks.

AI whose central capability is producing new text, code, images, audio, video, or structured data from a learned generative model.

Executing a trained or otherwise specified model on input to produce scores, predictions, decisions, or generated content.

A target or reference value attached to an example for supervised learning, evaluation, or later analysis.

A high-capacity learned model trained on large token sequences to estimate or generate language, often with related capabilities in code and other modalities.

A family of methods that adjusts a model from data or experience so its performance improves against a defined objective.

The finite mapping between token types and identifiers that a tokenizer uses and a model's embedding and output layers are trained to interpret.

A model that learns or applies relationships across more than one modality, such as text, images, audio, video, sensor signals, or structured data.

An AI system designed, authorized, or validated for a bounded task and operating domain rather than broad transferable competence.

The field of building computational systems that analyze, retrieve, translate, or generate human language in text or speech.

A computational model of connected units whose learned parameters transform inputs into predictions, representations, or generated outputs.

A learned numerical value that helps determine how a trained model transforms inputs into outputs.

A rule that assigns probability to possible outcomes or ranges of outcomes, with total probability equal to one.

An attention formulation that compares queries with keys, normalizes the scores into weights, and applies those weights to corresponding values.

A contested claim that an AI system can improve its own capability in repeated rounds, where each round's gains make the next round more effective.

Estimating a numeric response or the relationship between a response and one or more inputs.

A family of methods in which an agent learns a policy by interacting with an environment and optimizing expected cumulative reward.

Learning in which supervisory targets are generated from the structure of otherwise unlabeled data, such as predicting hidden or next tokens.

Machine learning from labeled examples that pair inputs with desired outputs.

Artificially generated records designed to reproduce selected properties of real or imagined data for training, testing, simulation, or controlled sharing.

One vocabulary unit produced or consumed by a model, such as a word, subword, byte sequence, punctuation mark, special marker, or non-text unit.

The extra token count, cost, latency, or lost context capacity imposed when a tokenizer represents equivalent content less efficiently in one language, script, domain, or notation than another.

An algorithm, vocabulary, and set of text-processing rules that encode input into token identifiers and decode identifiers back into usable output.

The process of adjusting a model's parameters to reduce a defined objective on training data, with the aim of producing useful behavior on new inputs.

A neural-network architecture that combines self-attention, position information, and feed-forward transformations to model relationships within sequences or other structured inputs.

An imitation game in which a human judge uses text conversation to assess whether a machine can be distinguished from a human participant.

Learning patterns, structure, or representations from data without supplied target labels.

A function that transforms a neural network unit's pre-activation, usually to add nonlinearity or control how much signal passes forward.

A training loop in which the learner chooses which unlabeled examples should be sent to an oracle for labels.

A model trained to encode an input into an internal representation and decode that representation into a reconstruction of the input.

A reverse pass through a differentiable computation graph that computes how the loss changes with respect to each parameter.

A language model trained to predict each token from the tokens that precede it, without access to future positions.

The processing work or capacity used to train and run an AI system, measured with quantities such as operations, accelerator time, throughput, or cost.

A representation-learning approach that trains examples judged related to be closer and examples judged different to be farther apart in a learned space.

A neural network that applies shared learned filters across local regions of spatial or sequential data.

A loss that measures how poorly a predicted probability distribution represents the target distribution.

Expanding or varying training examples through transformations or generation intended to preserve task-relevant meaning.

A model component that turns an internal representation or prior outputs into a reconstruction, prediction, or target sequence.

A Transformer-style sequence model that uses causal self-attention to predict the next token without a separate encoder stack.

A generative model that learns to reverse a gradual noising process, producing samples by repeatedly denoising an initial random state.

A preference-training method that increases a model's relative likelihood of chosen responses over rejected responses without training a separate reward model.

A model component that turns an input into an internal representation for classification, retrieval, reconstruction, or generation.

A model with separate paths for representing a source input and generating a target output conditioned on that representation.

One nominal pass through a training dataset, usually divided into batches and followed by another shuffle or sampling cycle.

Training a shared model from updates computed by distributed participants while their raw training data remains local.

An additional training phase that adapts a pretrained model by updating all parameters or a selected subset on new data.

A general-purpose AI model at or near the leading edge of broadly evaluated capability at a particular time.

The ability of a learned model or system to perform well on relevant data, tasks, or environments that were not used to fit it.

A generative model trained as a game between a generator that produces candidates and a discriminator that tries to detect generated samples.

An iterative optimization method that moves parameters opposite the gradient of an objective being minimized.

A throughput-oriented parallel processor widely used for neural-network training and inference.

A configuration choice that governs a model or learning procedure but is not ordinarily learned as a model weight.

Fine-tuning a pretrained model on instruction-and-response examples so it learns to perform tasks described in natural language.

A space of unobserved variables or learned representations used to model structure behind observed data.

A hyperparameter that sets the base step size for parameter updates during optimization.

A function that converts a model's prediction and target into a training penalty for an example or batch.

A parameter-efficient fine-tuning method that freezes selected base weights and learns low-rank matrices representing their updates.

A language model trained to reconstruct selected hidden tokens from the visible context on both sides.

A model architecture in which a learned router selects one or more expert subnetworks for each input or token.

A saved snapshot of model parameters and, when needed, the surrounding state required to resume training.

Training a student model to reproduce useful behavior from a teacher model or ensemble, often to reduce inference cost.

The learned numerical values in a model's connections or transformations, often used loosely for all learned parameters.

The mathematical quantity an optimization or search process is configured to minimize or maximize.

An AI system released with the rights and preferred form needed to use, study, modify, and share the system.

A model whose trained parameters are distributed under stated terms, without implying that its training process or complete system is open source.

An algorithm that turns gradients and accumulated training state into parameter updates.

A failure to generalize in which performance on development data is better than performance on representative unseen data.

A numerical signal that lets a sequence model distinguish token positions or distances.

Training performed before a later adaptation stage, usually to learn reusable representations or capabilities from a broader dataset.

A model whose owner materially controls access to its weights, development artifacts, operation, modification, or redistribution.

Mapping model values to a lower-precision representation to reduce storage, memory traffic, or computation.

A fine-tuning method that backpropagates through a frozen quantized base model into trainable LoRA adapters.

A neural-network architecture that applies a recurrent update while carrying hidden state through a sequence.

A change to the learning problem intended to improve performance on unseen data, often by limiting or perturbing what the model can fit.

A training approach that learns a reward signal from human judgments and uses reinforcement learning to optimize a model against that signal.

Empirical relationships fitted to how model loss or performance changes with resources such as parameters, data, or compute.

A language model small enough for a target deployment's memory, latency, cost, or control constraints relative to contemporary large models.

A group of training examples processed together to estimate a gradient before an optimizer update.

The examples and signals used to fit a model's learned parameters during pretraining, fine-tuning, or other learning procedures.

Using representations or parameters learned in a source task or domain to help a target task or domain.

A latent-variable generative model trained with variational inference so it can reconstruct data and sample new examples.

A bounded search algorithm that keeps a fixed number of high-scoring partial sequences at each decoding step.

Prompting a model with worked intermediate reasoning, or asking it to generate intermediate steps, before producing an answer.

Designing how instructions, state, knowledge, examples, tools, and feedback are selected, structured, and delivered to a model at the moment they are needed.

An inference scheduling technique that adds and removes generation requests at iteration boundaries as capacity becomes available.

Supplying a small set of worked examples in context to steer task behavior without updating model weights.

A model interface in which the model selects a named function and supplies structured arguments for application code to execute.

Generating each next token by selecting the current highest-probability candidate.

Connecting an AI output to identifiable evidence, data, observations, or constraints that support its claims.

Generated content that is unsupported by the required source or context, conflicts with it, or presents invented material as factual.

A model's ability to adapt behavior from instructions, examples, or patterns supplied within the current context without parameter updates.

A token supplied to a model for an inference call, including user content and any instructions, history, retrieved material, tool definitions, or other context assembled by the system.

Stored attention keys and values from earlier tokens that an autoregressive transformer reuses instead of recomputing them for every new token.

An unnormalized model score supplied to a normalization function such as softmax; in statistics, logit also means the logarithm of odds.

A request-level ceiling on the tokens a model may generate, which can include hidden reasoning as well as visible output on some APIs.

Supplying one worked example in context to demonstrate the desired task or output pattern.

A token generated by a model as part of its response, often metered separately from input tokens.

Input supplied to a model to condition the output, including instructions, examples, context, and user data.

Connecting multiple model calls so one stage's result becomes context or input for a later stage.

Designing and testing model inputs to elicit useful behavior from a particular model and task.

A model optimized to spend additional inference effort on multi-step problem solving before returning an answer or action.

A provider-reported token used by a reasoning model for intermediate inference work before or alongside its visible answer.

Selecting an output token from the probability distribution produced by a generative model.

A configured token or text pattern that causes generation to terminate when produced.

Model output constrained to a machine-readable schema such as JSON Schema so downstream software can validate and consume it reliably.

A high-authority runtime instruction layer that sets a model's role, constraints, tools, and operating context before it handles user input.

A decoding setting that rescales token logits before sampling, making the next-token distribution sharper or flatter.

Computation spent after a request arrives on reasoning, search, candidate generation, tool use, or verification before returning an answer.

A decoding method that samples only from the smallest set of candidate tokens whose cumulative probability reaches a chosen threshold.

Recognizing unseen classes through shared semantics, or, in current LLM usage, attempting a task without demonstrations in the prompt.

A system for capturing, retaining, selecting, and delivering prior state so it can influence an agent's later decisions.

Splitting documents or data into retrieval units that can be indexed, selected, and placed into model context.

The degradation in an agent's ability to find, follow, or correctly weigh relevant information as its accumulated context becomes longer, noisier, stale, repetitive, or internally conflicting.

Agent state intentionally retained across runs and made available later, with explicit ownership, provenance, retention, and correction rules.

A family of retrieval-augmented generation methods that builds or uses graph structure to retrieve relationships, communities, paths, or corpus-level themes.

A graph-shaped representation of entities or concepts and their named relationships, often governed by a schema and linked to supporting sources.

Reusing model computation for an unchanged prompt prefix or context block across requests to reduce repeated input processing, latency, and cost.

Reducing the tokens sent to a model while attempting to preserve the instructions, evidence, state, and relationships needed for the task.

Rescoring an initial candidate set with a second model or rule to improve its ordering before selection.

Generating a response after retrieving relevant material from an external knowledge source and adding it to model context.

Retrieval that compares learned representations of a query and candidate items instead of relying only on exact term overlap.

A data system that stores vectors with identifiers and metadata and supports similarity search over them.

The bounded task state actively delivered to an agent during a run, including current goals, constraints, observations, and pending work.

A software system in which a model interprets a goal or input, decides among actions, uses tools or other capabilities, observes results, and continues until completion, handoff, or termination.

The software layer that surrounds a model with instructions, context assembly, tools, state, permissions, control flow, budgets, verification, observability, and recovery.

The repeated cycle in which an agent observes state, selects an action, invokes a tool or model, receives feedback, updates state, and decides whether to continue.

A task-specific arrangement of prompts, tools, control logic, and feedback wrapped around a model to improve performance.

A versioned package of instructions, procedures, examples, and sometimes code or resources that teaches an agent how to perform a repeatable class of work.

The ordered record of states, model outputs, actions, tool results, and transitions produced during an agent run.

Describing a system that can choose and sequence actions toward a goal with some runtime discretion rather than only produce a single predetermined response.

The degree to which a system can select and execute actions without case-by-case human direction or approval.

A conversational software interface that accepts natural-language input and returns responses, whether powered by rules, retrieval, generative models, or combinations of them.

A directed representation of the steps an agent system may execute and the conditions that select what runs next.

A directed graph with no path that returns to an earlier node.

A run-oriented graph of executable steps and the control or data dependencies connecting them.

Designing an agent system as explicit nodes, state, and transitions so deterministic control and model judgment have visible boundaries.

A transfer of active responsibility, context, and next-action authority from one agent or person to another.

An operating pattern in which a human participates directly in the decision or execution path before work can continue.

An operating pattern in which a system acts while a human supervises, receives evidence or alerts, and can intervene or stop it.

An operating condition in which a system completes a scoped activity without routine human participation in its action path.

Designing the feedback cycles around an agent so action, verification, operational triggers, and system improvement have explicit state, evidence, limits, and stopping conditions.

A system in which multiple agents communicate, specialize, coordinate, compete, or verify one another to accomplish work.

Coordinating tasks, agents, tools, state, dependencies, budgets, failures, and lifecycle across a workflow.

The component or role that admits work, assigns it, coordinates dependencies and concurrency, tracks state, handles retries, and determines handoffs or completion.

A model of a system as explicit states and permitted transitions triggered by events or conditions.

An agent invoked by another agent or orchestrator to perform a bounded portion of a larger task.

Breaking an objective into smaller units with explicit dependencies, interfaces, acceptance criteria, and ownership.

A defined sequence or graph of activities, decisions, states, and transitions used to achieve an outcome.

An A2A discovery document that describes an agent's identity, supported interfaces, capabilities, skills, and security schemes.

An open protocol for discovering and collaborating with independent agent services through messages, tasks, artifacts, and updates.

A documented interface through which one software component requests data or behavior from another.

A bounded action or class of action that a system component can perform, usually paired with explicit authority and scope.

A short-lived execution environment created for a task or run and destroyed when that work ends.

A reconstructable chain linking an outcome to the intent, identities, system versions, inputs, actions, evidence, and state changes that produced it.

Elapsed time between a defined starting event and a defined response or completion point.

The MCP initialization exchange in which a client and server declare the optional protocol features they support.

The MCP protocol component created by a host to establish and manage a connection with one server.

An intermediary that fronts one or more MCP servers and centralizes routing, authentication, policy, limits, or audit.

The AI application that coordinates users, models, policy, and one or more MCP server connections.

A reusable message template exposed by an MCP server for a user to select and supply to a model interaction.

Data exposed by an MCP server under a URI so a host can retrieve it and decide whether to place it in model context.

A program or service that exposes tools, resources, prompts, or other negotiated MCP capabilities to a connected client.

An executable operation exposed by an MCP server for model discovery and invocation through a defined schema.

An open client-server protocol for connecting AI applications to external tools and context through standard discovery and interaction primitives.

A constraint on the requests, tokens, compute, or actions allowed within a defined time window.

An execution environment that enforces boundaries around files, processes, networks, credentials, and other resources.

The amount of defined work a system completes during a unit of time.

An executable capability an AI application or agent can invoke to retrieve information, run computation, or change state.

A randomized controlled experiment that exposes comparable groups to different variants and compares a predefined outcome.

An experiment that removes or changes one component while holding others as constant as practical to estimate that component's contribution.

An explicit outcome in which a model or evaluator declines to answer, act, or judge because evidence or authority is insufficient.

Explicit conditions an outcome must satisfy before it can be accepted, promoted, or declared complete.

The proportion of evaluated predictions counted as correct under a stated label set, unit of analysis, and decision rule.

A defined package of tasks, data, metrics, and execution rules used to compare systems under the same evaluation regime.

The degree to which a system's stated probabilities match observed outcome frequencies across comparable predictions.

A change over time in the relationship between inputs and the correct target or decision.

A table counting predicted classes against actual classes, including true and false positives and negatives.

A resampling method that estimates generalization by repeatedly training and evaluating on different non-overlapping subsets of available data.

A change over time in the distribution of system inputs or features.

A capability or behavior that appears qualitatively new at a larger scale or higher level of system interaction rather than as an obvious continuation of smaller-scale measurements.

A systematic measurement of model or system behavior against defined tasks, criteria, datasets, or operational outcomes.

The harmonic mean of precision and recall.

An outcome incorrectly classified as absent when it is actually present.

A negative case that a classifier or detector incorrectly labels as positive.

A curated set of reference examples with reviewed expected outcomes used for evaluation or regression testing.

A deterministic check, model, or human protocol that judges an output or trajectory against stated criteria.

Examples kept separate from training and development decisions so evaluation measures generalization to data the workflow has not optimized against.

Checking an outcome with evidence, components, context, or authorities meaningfully separated from the system that produced it.

Using a language model to evaluate, compare, classify, or score outputs produced by models or agents.

A broad operational term for model behavior or performance changing relative to an accepted baseline.

The property that identical-looking requests can produce different behavior because of sampling, concurrency, infrastructure, model updates, or hidden state.

The ability to infer a system's internal state and behavior from emitted traces, logs, metrics, events, and artifacts.

A mechanism that can determine the expected or acceptable result for a task, such as a compiler, formal specification, invariant, test suite, or reference implementation.

Among predicted-positive cases, the proportion that are truly positive.

Among truly positive cases, the proportion correctly identified as positive.

Achieving a high measured reward through behavior that exploits the metric or evaluator without accomplishing the intended objective.

Satisfying the literal specification or metric in a way that violates its intended purpose.

A failure mode in which a model favors agreement with a user's stated belief or preference over an independently supported answer.

A captured sequence of model calls, tool calls, events, timings, state changes, and outputs from an execution.

A control point that blocks promotion until required evidence has been produced and validated.

A repeated execute, observe, compare, and correct cycle that withholds completion until an attempted result satisfies explicit evidence or acceptance criteria.

The effort to keep an AI system's behavior compatible with intended human goals, constraints, and values across relevant conditions.

A systematic tendency in data, modeling, or decision processes that skews estimates, errors, treatment, or impacts.

The normative and technical treatment of how an AI system distributes errors, benefits, burdens, and opportunities across people or groups.

Structured adversarial testing intended to discover ways an AI system can fail, be misused, or violate constraints.

The field and practice of keeping AI-system risks within acceptable bounds across design, deployment, operation, and retirement.

The people, artifacts, services, and infrastructure whose integrity can change an AI system's behavior or authority.

Data demonstrably fit for a named AI use, with the quality, structure, rights, provenance, coverage, and separation that use requires.

An input constructed to make a model produce an incorrect or attacker-chosen result, often through a change that appears insignificant to a person.

Training on adversarially generated examples to improve robustness against attacks inside a defined threat model.

Attributing human mental states, motives, understanding, or emotion to an AI system because its behavior resembles a person's.

A structured argument, supported by evidence, that a system is acceptably safe or dependable within a stated scope.

The plausible scope of damage, exposure, or irreversible change if a component, credential, or action fails or is compromised.

Corrupting information placed into an agent's active or persistent context so later decisions are based on false facts, malicious instructions, or distorted state.

Introducing malicious, misleading, or strategically biased data into training, tuning, retrieval, memory, or evaluation pipelines to alter later behavior.

A mathematical privacy framework that bounds how much a computation's output can change because one person's data is included or removed.

Prompt injection delivered directly through the current user's message or another explicit input channel.

Granting an AI system more functionality, permissions, autonomy, or action scope than required for its intended task.

The ability to provide a human-usable account of why a system produced a particular output or action.

A rule, model, filter, policy check, or enforcement mechanism intended to constrain inputs, outputs, or actions of an AI system.

Malicious instructions embedded in external content such as webpages, documents, email, code, tool results, or retrieved memory that an AI system later processes.

The degree to which a human can understand how a model represents information or produces behavior.

An input strategy intended to make a model bypass or disregard its trained or instructed safety restrictions.

Granting an identity or component only the minimum permissions needed for a bounded task, for no longer than needed.

Information that can identify, distinguish, or be linked to a specific person, alone or in combination with other data.

Expressing policy in machine-readable rules that can be versioned, tested, reviewed, and enforced by software.

An explanation produced after a model has generated a prediction or action, often by analyzing inputs, outputs, or a separate approximation.

Manipulating an AI system by placing instructions in input or data that the model treats as authoritative enough to alter intended behavior.

Evidence describing the origin, ownership, custody, transformation, and version history of data, code, models, skills, tools, or claims.

Granting different levels of agent discretion and authority according to task verifiability, reversibility, sensitivity, and blast radius.

Restoring a previously known-good software, configuration, model, policy, or data state after a failed or harmful change.

Manipulating a tool's description, schema, implementation, or output so an agent selects unsafe actions or incorporates attacker-controlled instructions.

A contested label for low-quality, low-effort, often high-volume content produced or amplified with generative AI.

A repeatable operating cycle for building, testing, deploying, monitoring, improving, and governing an agent system.

A software-development method in which a coding agent plans and executes multi-step work while a human directs, reviews, or governs the outcome.

The discipline of designing software work, environments, and controls so AI agents can perform substantial engineering without displacing human product judgment and accountability.

A derogatory slang term for a robot, AI system, or automated technology, used jokingly or hostilely to express disdain for machines or their perceived replacement of human work.

An AI agent that can inspect a software project, change it through development tools, evaluate the result, and iterate toward a software outcome.

An AI system that helps a human understand or change software while the human remains the primary driver of the workflow.

The accumulated loss of shared human understanding, reasoning continuity, or recovery competence caused by repeatedly delegating cognition without rebuilding comprehension.

The way software and repository structure changes the amount, quality, and retrieval cost of context an agent needs for a correct change.

A bounded process for improving an agent's prompts, skills, memory, workflows, routing, or harness under independent evaluation and reversible rollout.

The full cost of producing, validating, correcting, and operating work divided by outcomes that pass acceptance and remain useful for a defined period.

A domain-bounded software production system in which humans specify intent, risk, and policy while a model-harness-environment system plans, builds, verifies, ships, observes, and repairs software with little routine human intervention.

A living digital representation of a specific system or environment, connected closely enough to test or reason about its real behavior.

The finite supply of skilled human judgment available to specify, review, govern, and recover automated work.

A proposed control map showing how optimizers, evaluators, counter-metrics, auditors, and promotion gates govern changes to an AI system.

The operational discipline for evaluating, deploying, observing, governing, and maintaining applications built around large language models.

The engineering and operational practices used to build, deploy, observe, govern, and maintain machine-learning systems throughout their lifecycle.

An emerging label for optimizing an AI workflow around accepted, durable results instead of easy-to-count activity.

Evidence from sustained real operation used to judge whether a system keeps producing acceptable outcomes after launch.

The governed decision to move an artifact or change into a more trusted lifecycle state after required evidence and policy checks.

A machine-readable, human-auditable agreement defining one agent run's objective, authority, evidence, limits, and recovery path.

A failure that looks mechanically successful while producing the wrong meaning, intent, binding, or real-world consequence.

A control pattern in which all changes to a sensitive persistent state pass through one authorized writer.

A repeatable production system that turns software demand into accepted, operated software through reusable processes, tooling, controls, and feedback.

A development approach in which an explicit, versioned specification materially guides implementation and verification.

An explicit allocation or ceiling for model-token consumption across a named scope, with rules for warning, stopping, and exceptions.

The amount or rate of model-token consumption across a defined unit of work and time window.

The relationship between quality-constrained outcomes and the model tokens consumed to produce them.

Deliberately or incentive-drivenly maximizing the tokens consumed by AI work, often by expanding context, reasoning, turns, agents, or tasks, while treating greater usage as a route to capability or a proxy for productivity.

An emerging counterterm for systematically reducing AI token consumption while preserving an explicit threshold for useful outcome quality.

Token-consuming AI activity that produces insufficient learning, accepted work, or maintained value for its total cost.

Sarah Friar's proposed AI value scorecard connecting useful work, successful-task cost, dependability, and value at scale.

A loose AI-assisted programming style that accepts generated behavior with little inspection or understanding of the underlying code.