Glossary

4D Panoptic Segmentation is a temporal LiDAR scene-understanding task that assigns every evaluated 3...

A2A Protocol (Agent-to-Agent Protocol) is an open standard for discovery, messaging, task lifecycle ...

Accumulated Local Effects (ALE) is a global model-inspection method that averages prediction differe...

ACP (Agent Client Protocol) is an open JSON-RPC protocol for communication between a code editor or ...

An Acquisition Function is a comparatively cheap decision rule that assigns utility to one or more c...

Activation Patching is a causal intervention method that replaces selected internal activations in o...

Activation Steering is an inference-time intervention that modifies a model's internal activations a...

Active Learning is a supervised-learning process in which a learner chooses which unlabeled examples...

Adapter is base-dependent trainable state that changes a pretrained model's behavior while most base...

Adaptive MCMC is a class of sampling algorithms that updates parameters of a Markov transition using...

Adjusted Rand Index is a symmetric measure of agreement between two hard partitions that counts whet...

AES (Advanced Encryption Standard) is a symmetric block cipher algorithm adopted by the U.S. governm...

Agent Development Kit is a software framework that packages reusable interfaces and runtime services...

Agent Harness is the application control system around an AI agent that assembles context, constrain...

Agent Memory is the governed storage and retrieval of information that an AI agent may use across st...

Agent Runtime is the execution environment that creates, schedules, isolates, advances, suspends, re...

Agent Trajectory is a structured, causally ordered record of the observable decisions, messages, too...

An Agentic Workflow is a business or technical process that delegates selected runtime decisions to ...

AGI (Artificial General Intelligence) is a contested concept for an AI system with broad, transferab...

AI Accelerator is specialized hardware and its supporting software stack designed to execute machine...

AI Agent is a software system in which an AI model can choose the next step, tool, or handoff from o...

AI App Builder is a category of no-code/low-code platforms that use large language models to generat...

AI Code Review is the use of machine-learning or language-model systems to inspect proposed code cha...

AI Guardrails are versioned technical and operational controls that detect, constrain, block, transf...

An AI Watermark is a signal intentionally embedded in generated or manipulated content so a correspo...

Aleatoric Uncertainty is uncertainty attributed to variability in an outcome that remains unresolved...

Algorithmic Fairness is the sociotechnical practice of defining, measuring, and governing how an aut...

Annealed Importance Sampling is a Monte Carlo method that estimates a ratio of normalizing constants...

Answer Correctness is an evaluation property that measures whether the factual claims or task result...

Answer Faithfulness is a RAG evaluation property that measures the proportion of factual answer clai...

Answer Relevance is an evaluation property that measures how directly and sufficiently a generated r...

API (Application Programming Interface) is a set of rules, protocols, and tools that allows differen...

An API Gateway is a server component that acts as the single entry point for all client API requests...

Approval Gate is a deterministic enforcement checkpoint that blocks a proposed AI action until trust...

Artificial Intelligence (AI) is the field and set of techniques for building machine-based systems t...

ASCII (American Standard Code for Information Interchange) is a character encoding standard that ass...

Aspect Ratio is the proportional relationship between the width and height of an image, video, or di...

Attention Mechanism is a differentiable operation that scores permitted Query-Key pairs, normalizes ...

Attribution Patching is a first-order approximation to activation patching that estimates how replac...

AutoGen is Microsoft's maintenance-mode framework for conversational agents and event-driven multi-a...

Automated Circuit Discovery (ACDC) is an intervention-based mechanistic interpretability algorithm t...

Autonomous Driving is a technology that enables vehicles to navigate and operate without human inter...

Average Multi-Object Tracking Accuracy is a confidence-swept multi-object tracking score that averag...

Average Multi-Object Tracking Precision is a confidence-swept localization measure that averages the...

Average Tracking Accuracy is a VACE metric that globally matches ground-truth and predicted tracks b...

Average Treatment Effect is the mean causal contrast in a declared target population between each un...

AWQ (Activation-aware Weight Quantization) is a post-training, weight-only quantization method that ...

A Backdoor Attack is an adversarial manipulation that implants hidden behavior in a machine learning...

Backpropagation is the reverse-mode gradient computation commonly used for neural network objectives...

Balanced Accuracy is the unweighted mean of Recall calculated for every declared class, giving each ...

Barcode is a machine-readable representation of data using patterns of parallel lines (1D) or geomet...

Base64 is a binary-to-text encoding scheme that represents binary data in an ASCII string format by ...

A Bayesian Neural Network is a neural network model that places a prior distribution over selected p...

Bayesian Optimization is a sequential, model-based method for optimizing an expensive black-box obje...

Bayesian Quadrature is a probabilistic numerical-integration method that places a prior, commonly a ...

Bearer Token is an access token type used in HTTP authentication where the client presents a token t...

Bi-Encoder is a model architecture that encodes two inputs independently into reusable vectors and c...

Binary (also known as base-2) is a numeral system that uses only two symbols: 0 and 1. It is the fun...

BM25 is a family of probabilistic lexical ranking functions that scores a document from query-term e...

BMI is a numerical value calculated from a person's weight and height, used as a screening tool to c...

Brier Score is a strictly proper scoring rule that averages the squared distance between a predicted...

Calinski-Harabasz Index is an internal cluster-validity measure that divides between-cluster dispers...

Callback Handler is a framework hook that receives typed lifecycle events around model, tool, retrie...

Cardinality-Balanced Multi-Target Multi-Bernoulli Filter is a Random Finite Set approximation that r...

Cardinalized Probability Hypothesis Density Filter is a Random Finite Set approximation that propaga...

Catastrophic Forgetting is a substantial decline in a model's performance on previously learned task...

Causal DAG is a Directed Acyclic Graph whose nodes represent causally defined variables and whose ar...

Causal Discovery is the task of learning identifiable features of causal structure from data under e...

Causal Forest is a forest-based estimator that builds adaptive covariate neighborhoods for estimatin...

Causal Inference is the process of defining and estimating how an outcome would change under a speci...

Causal Mediation Analysis studies how much of an intervention's effect operates through a specified ...

Causal Meta-Learner is a framework that combines ordinary predictive learners through a treatment-aw...

Causal Scrubbing is a mechanistic interpretability method that converts an explicit computational hy...

Causal Transportability is the identification and estimation of a causal effect in a target populati...

CDN is a geographically distributed network of proxy servers and data centers that work together to ...

Centered Kernel Alignment (CKA) is a normalized similarity measure between two centered Gram matrice...

Chain Orchestration is an LLM application pattern in which code defines an ordered path of model, re...

Chain-of-Thought (CoT) is a prompting technique that elicits intermediate natural-language steps bef...

Chat Template is a model-specific serialization contract that converts structured messages, roles, t...

Chatbot is a software system that exchanges conversational turns with a user through text, voice, or...

ChatModel is a framework-level adapter interface that accepts structured conversational requests and...

Checksum is a small-sized block of data derived from another block of digital data for the purpose o...

Chunk Overlap is the repeated text shared between adjacent document chunks so that information near ...

Chunk Size is the token, character, or structural length chosen for each document unit indexed in a ...

Chunking is the process of splitting long documents or data sources into smaller retrievable units t...

Citation is a verifiable mapping from an AI-generated claim to a specific source span, record, or to...

Classification Accuracy is the fraction of evaluated examples whose predicted hard label exactly mat...

CMYK is a subtractive color model used in color printing, where cyan, magenta, yellow, and black ink...

CNN (Convolutional Neural Network) is a neural-network architecture that applies shared local kernel...

Cold Start is the startup interval triggered when ready model-serving capacity is unavailable, measu...

Computer Vision is the field that develops computational methods for extracting, estimating, or gene...

Concept Bottleneck Model (CBM) is a predictive architecture that maps inputs to a declared set of hu...

Concept Drift is a time-indexed change in the conditional target relationship `P_t(Y|X)`, so the map...

Concept Whitening is an intrinsic interpretability technique that replaces a normalization layer wit...

Conditional Average Treatment Effect is the expected difference between Potential Outcomes under tre...

Confidence Calibration is the property and process by which a model's stated probabilities are made ...

Conformal Prediction is a model-agnostic statistical framework that converts point predictions or mo...

Confounding is distortion of a treatment-outcome association caused by differences in common causes ...

Confusion Matrix is a table whose rows and columns represent actual and predicted classes, so each e...

Constrained Decoding is an inference technique that filters an LLM's allowed next tokens so the gene...

Context Budget is an application-level policy that allocates a model request's finite token capacity...

Context Caching is a cross-request inference optimization that reuses provider- or runtime-managed s...

Context Compression is a lossy transformation that reduces the context supplied to an LLM while pres...

Context Engineering is the design and operation of the information supplied to a language model for ...

Context Precision is a ranking-aware RAG evaluation metric that rewards relevant retrieved contexts ...

Context Recall is a RAG evaluation metric that measures the proportion of a declared required-eviden...

Context Window is the maximum tokenized sequence a model endpoint accepts for one inference request ...

A Contextual Bandit is a sequential decision problem in which a policy observes the current context,...

Continuous Batching is an LLM serving scheduling pattern that can admit, advance, pause, or retire r...

CORS (Cross-Origin Resource Sharing) is a security mechanism that allows web browsers to make reques...

Counterfactual Explanation is a local explanation that describes one or more constrained changes to ...

Counterfactual Fairness is a causal fairness criterion requiring a predictor's distribution for the ...

Covariance Intersection is a conservative data-fusion method that combines uncertain estimates witho...

Covariate Shift is a distribution-change assumption under which the input marginal changes from sour...

CrewAI is an open-source Python framework that organizes agent applications with Agents, Tasks, Crew...

Cron is a time-based job scheduler in Unix-like operating systems that enables users to schedule com...

Cross-Encoder is a pair-scoring model architecture that jointly encodes a query and candidate in one...

Cross-Fitting is a sample-splitting procedure that trains nuisance models on one subset and evaluate...

CSS (Cascading Style Sheets) is a stylesheet language used to describe the presentation of HTML and ...

CSV (Comma-Separated Values) is a plain text file format that stores tabular data using commas to se...

A Cubature Kalman Filter is a nonlinear Gaussian state estimator that approximates transformed momen...

Curriculum Learning is a training strategy that changes the distribution, weight, or availability of...

Data Poisoning is a training-time integrity or availability attack in which an adversary controls pa...

Data URL is a URI scheme that allows embedding small data items inline in web documents as if they w...

Dataset Curation is the governed process of selecting, transforming, labeling, partitioning, documen...

Davies-Bouldin Index is an internal cluster-validity measure that averages, for every cluster, its l...

DBSCAN is a density-based clustering algorithm that expands connected neighborhoods around core samp...

Decode Phase is the autoregressive LLM inference stage after Prefill in which accepted output tokens...

Deep Ensembles are collections of neural networks trained as distinct members for the same predictiv...

Deep Gaussian Process is a hierarchical probabilistic model that composes multiple Gaussian Process ...

Deep Kernel Learning (DKL) is a modeling approach that applies a valid kernel to representations pro...

Deep Learning is a family of machine-learning methods that composes multiple parameterized transform...

Demographic Parity is a group fairness criterion requiring a model prediction or downstream decision...

Dense Retrieval is a first-stage search method that encodes queries and documents into compact numer...

Diff is a comparison technique that identifies and displays the differences between two sets of data...

Difference-in-Differences is a quasi-experimental design that estimates a treatment effect by subtra...

Differential Privacy is a mathematical framework that bounds how much the probability of any output ...

Diffusion Maps is a nonlinear dimensionality-reduction method that converts sample affinities into a...

Diffusion Model is a family of generative models that learns a reverse denoising transition from noi...

Direct Logit Attribution is a mechanistic interpretability method that projects individual residual-...

Disaggregated Serving is an LLM inference architecture that runs prompt prefill and token decoding i...

Distillation (Knowledge Distillation) is a training method in which a student model learns from sign...

Distribution Shift is a change between the joint data-generating distribution used to develop or eva...

DiT (Diffusion Transformer) is a transformer-based denoising or velocity-prediction backbone that re...

DNS (Domain Name System) is a hierarchical and distributed naming system that translates human-reada...

Document Loader is a source-facing ingestion component that reads content and source metadata from f...

Document Transformer is a pipeline component that converts one or more loaded document objects into ...

Doubly Robust Estimation combines an outcome or reward model with an inverse-propensity correction s...

DPO (Direct Preference Optimization) is a preference-learning objective that adjusts a language mode...

DVPQ (Depth-Aware Video Panoptic Quality) is a joint metric that masks panoptic predictions at pixel...

DVPS (Depth-Aware Video Panoptic Segmentation) is a dense video-perception task that jointly predict...

Dynamic Treatment Regime is a sequence of prespecified decision rules that maps the information avai...

Effective Sample Size is a diagnostic that expresses how much information a weighted or dependent sa...

Eino is an open-source Go framework in the CloudWeGo ecosystem that provides typed LLM components, C...

Elliptical Slice Sampling is a Markov Chain Monte Carlo method for latent variables with a multivari...

Embedding is a learned numerical representation that maps text, images, entities, or other inputs in...

Emoji is a standardized set of pictographic symbols and icons encoded in Unicode, used in digital co...

Encryption is the process of converting plaintext data into an unreadable format (ciphertext) using ...

An Ensemble Kalman Filter is a sequential data-assimilation method that propagates an ensemble throu...

Epistemic Uncertainty is uncertainty about the predictive mechanism, parameters, representation, or ...

Epoch time refers to the reference point (January 1, 1970, 00:00:00 UTC) from which Unix timestamps ...

Equalized Odds is a group fairness criterion requiring a prediction to be statistically independent ...

Error-State Kalman Filter is an Extended Kalman Filter formulation that propagates a nominal state w...

Escape Character is a character that invokes an alternative interpretation of subsequent characters ...

EXIF is a standard that specifies the formats for images, sound, and ancillary tags used by digital ...

Expected Calibration Error is a binned estimator that summarizes the weighted absolute difference be...

Expected Improvement is an acquisition function that scores a candidate by the posterior expectation...

Explainable AI (XAI) is the field of methods and practices for producing evidence or representations...

An Extended Kalman Filter is a nonlinear state estimator that propagates a mean through nonlinear dy...

F1 Score is the harmonic mean of Precision and Recall for a declared positive class at a specific de...

Feature Visualization is a family of interpretability methods that characterize what a neural-networ...

Federated Learning is a distributed machine learning approach in which clients train on local data a...

Few-Shot Learning is a family of machine-learning settings in which a system must generalize to a ne...

Fine-tuning is the process of continuing optimization from a pretrained model checkpoint on a target...

A Fixed-Lag Smoother is a state estimator that delays each output by a chosen lag so measurements fr...

Function Calling is a model interface that returns a structured proposal naming a tool and its argum...

G-computation is a causal standardization method that estimates an outcome under each intervention b...

GAN (Generative Adversarial Network) is an implicit generative model trained as a two-player game in...

Gaussian Mixture Model is a probabilistic density model that represents observations as a weighted s...

A Gaussian Process is a stochastic process whose function values at every finite collection of input...

Gaussian Process Classification is a Bayesian classification model that places a Gaussian Process pr...

Gaussian Process Latent Variable Model is a probabilistic nonlinear dimensionality-reduction model t...

Gaussian Process State-Space Model is a probabilistic dynamical model that places a Gaussian Process...

Generalized Labeled Multi-Bernoulli Filter is a labeled Random Finite Set tracker whose posterior is...

Generalized Optimal Sub-Pattern Assignment Metric is an unnormalized distance between finite sets th...

Generative AI is a class of models and systems that learns a probability distribution over data and ...

Geohash is a geocoding system that encodes geographic coordinates (latitude and longitude) into a sh...

Gibbs Sampling is a Markov Chain Monte Carlo method that updates one variable or block at a time by ...

GIF (Graphics Interchange Format) is a bitmap image format developed by CompuServe in 1987 that supp...

Glob is a pattern matching syntax used to specify sets of filenames or paths using wildcard characte...

Go (also known as Golang) is a statically typed, compiled programming language designed at Google. I...

Golden Dataset is a versioned, human-reviewed evaluation set whose cases pair representative inputs ...

GPT (Generative Pre-trained Transformer) is OpenAI's family of generatively pre-trained Transformer ...

Grad-CAM (Gradient-weighted Class Activation Mapping) is a model-specific attribution method that us...

Gradient Descent is a first-order optimization method that iteratively updates parameters by subtrac...

A Gradient Inversion Attack is a privacy attack that uses gradients or model updates exposed during ...

Graph Orchestration is an execution pattern that represents an AI workflow as explicit nodes, transi...

GraphRAG (Graph Retrieval-Augmented Generation) is a RAG approach that represents entities and relat...

Grounding is the system practice of supplying a generative AI model with bounded external evidence a...

GRPO (Group Relative Policy Optimization) is an online policy-gradient method that estimates each sa...

GUID (Globally Unique Identifier) is Microsoft's implementation of the UUID standard, a 128-bit iden...

Gzip is a file compression format and software application that uses the DEFLATE algorithm to reduce...

Hallucination is confidently stated generated content that is erroneous or false relative to a decla...

Hamiltonian Monte Carlo is a Markov Chain Monte Carlo method that augments continuous parameters wit...

Hash is a mathematical function that converts input data of any size into a fixed-size string of cha...

Hessian Eigenmaps is a nonlinear manifold-learning method that estimates second derivatives in local...

Heterogeneous Treatment Effects are systematic differences in a treatment's causal effect across cov...

Hex Color is a six-digit alphanumeric representation of colors used in web design and digital graphi...

Hexadecimal (also known as hex or base-16) is a positional numeral system that uses 16 distinct symb...

Hierarchical Clustering is a family of algorithms that organizes samples into a nested tree by repea...

Higher Order Tracking Accuracy is a multi-object tracking score that combines detection accuracy and...

Hinge Loss is a convex margin-based classification loss defined in the binary case as `max(0, 1-yf(x...

HMAC (Hash-based Message Authentication Code) is a cryptographic technique that combines a secret ke...

HSIC (Hilbert-Schmidt Independence Criterion) is a kernel dependence measure defined as the squared ...

HSL Color (Hue, Saturation, Lightness) is a cylindrical color model that represents colors in a way ...

HTML (HyperText Markup Language) is the standard markup language for creating web pages and web appl...

HTML Entity (HTML Character Entity) is a string that begins with an ampersand (&) and ends with a se...

HTTP (Hypertext Transfer Protocol) is the foundation protocol of the World Wide Web, defining how me...

HTTPS (Hypertext Transfer Protocol Secure) is the secure version of HTTP that encrypts all communica...

Human-in-the-Loop is a control pattern in which an AI system pauses at defined decision points so an...

Hybrid Search is a retrieval architecture that runs two or more complementary candidate generators, ...

HyDE is a zero-shot retrieval method that generates one or more hypothetical answer-like documents, ...

Identity F1 Score is the harmonic mean of identification precision and identification recall after a...

Image Compression is the process of reducing the file size of digital images by removing redundant o...

Importance Sampling is a change-of-measure method that estimates an expectation under a target distr...

In-Context Learning (ICL) is inference-time behavior in which a language model uses instructions, de...

Indexer is a pipeline component that converts processed documents into durable retrieval records and...

Individual Conditional Expectation (ICE) is a model-inspection method that varies one selected featu...

Inducing Points are selected locations whose latent function values serve as a compact set of induci...

Inference (model inference) is the execution of a trained model on new input to produce a prediction...

Instruction Tuning is a supervised fine-tuning approach that trains a pretrained model on tasks expr...

Instrumental Variables are variables that shift treatment uptake, are independent of causes of the o...

Integrated Gradients is a feature-attribution method that multiplies each input-to-baseline differen...

Interacting Multiple Model is an approximate Bayesian state-estimation algorithm that mixes a fixed ...

Intrinsic Dimension Estimation is the task of inferring how many local degrees of freedom are needed...

An Invariant Extended Kalman Filter is a symmetry-aware state estimator that defines estimation erro...

Inverse Probability of Censoring Weighting is a method that gives each still-observed unit a weight ...

Inverse Propensity Scoring is an importance-weighting estimator that evaluates a target policy or tr...

IP Address (Internet Protocol Address) is a unique numerical identifier assigned to every device con...

ISO 8601 is an international standard for representing dates and times in a clear, unambiguous forma...

Isomap (Isometric Mapping) is a nonlinear dimensionality-reduction method that estimates manifold ge...

An Iterated Extended Kalman Filter is an Extended Kalman Filter variant that repeatedly relinearizes...

J&F Score is the arithmetic mean of DAVIS region similarity J and contour accuracy F, combining mask...

Jailbreak is an adversarial attempt to make an AI model or model-backed system violate a specified s...

JavaScript is a high-level, interpreted programming language that enables interactive web pages and ...

Joint Probabilistic Data Association Filter is a multi-target tracking approximation that assigns pr...

JPEG (Joint Photographic Experts Group) is a commonly used lossy compression method for digital imag...

JSON (JavaScript Object Notation) is a lightweight, text-based data interchange format that is easy ...

JSON Mode is a provider-specific LLM output setting that makes a normally completed response syntact...

JSON Schema is a vocabulary that allows you to annotate and validate JSON documents. It describes th...

JSONPath is a query language for JSON documents that enables extracting and filtering data from comp...

JWT (JSON Web Token) is a compact, URL-safe means of representing claims to be transferred between t...

K-Means Clustering is a partitioning algorithm that assigns each vector to one of `k` centroids and ...

A Kalman Filter is a recursive Bayesian estimator that computes the filtering mean and covariance fo...

Kernel Mean Embedding (KME) represents a probability distribution by the expected feature map of its...

Kernel PCA (Kernel Principal Component Analysis) is a nonlinear dimensionality-reduction method that...

Kernel Ridge Regression (KRR) is a supervised regression method that minimizes squared prediction er...

Kernel Trick is a computational technique that evaluates inner products in an implicit feature space...

Knowledge Graph is a connected representation of entities, relationships, and claims, often with exp...

KTO (Kahneman-Tversky Optimization) is an offline LLM alignment objective that learns from individua...

KV Cache (Key-Value Cache) is per-layer inference state that stores the attention Key and Value tens...

Label Shift is a distribution-change assumption under which class prevalence changes from source to ...

Labeled Multi-Bernoulli Filter is a multi-object tracker that represents each potential object by a ...

LangGraph is a stable 1.x low-level orchestration runtime for long-running, stateful workflows that ...

The Laplace Approximation is a local method that approximates a smooth probability distribution with...

Laplacian Eigenmaps is a nonlinear dimensionality-reduction method that builds a weighted neighborho...

Latency is the elapsed wall-clock time between two explicitly named events in an AI system, such as ...

LIME (Local Interpretable Model-agnostic Explanations) is a post-hoc method that explains one black-...

Linear Probing is an analysis method that trains a linear classifier or regressor on frozen internal...

Lint is a static code analysis warning or error generated by a linter to flag style violations, susp...

Linting is the process of using automated tools to analyze source code for potential errors, bugs, s...

An LLM (Large Language Model) is a parameterized neural model that estimates probabilities over toke...

LLM-as-Judge is a probabilistic evaluation method in which a language model assesses one or more can...

LLMOps (Large Language Model Operations) is the engineering and governance discipline for developing...

Local Tangent Space Alignment (LTSA) is a nonlinear dimensionality-reduction method that estimates t...

Local Tracking Metrics are horizon-parameterized multi-object tracking measures that recompute optim...

Locally Linear Embedding (LLE) is a nonlinear dimensionality-reduction method that reconstructs each...

Log Loss is the mean negative logarithm of the probability a classifier assigns to each observed lab...

The Logit Lens is a training-free interpretability method that applies a language model's final norm...

LoRA (Low-Rank Adaptation) is a parameter-efficient adaptation method that keeps selected pretrained...

LoRA Rank is the inner dimension r of the two trainable matrices used to represent a LoRA weight upd...

Lorem Ipsum is a pseudo-Latin placeholder text commonly used in graphic design, web development, and...

Lost in the Middle is a position-sensitive long-context failure mode in which a language model uses ...

LSTQ is a point-centric 4D LiDAR panoptic segmentation metric that combines semantic classification ...

Machine Learning (ML) is a field of artificial intelligence in which an algorithm fits a model from ...

Machine Unlearning is the process of updating a trained machine learning system so the influence of ...

Manifold Learning is a family of nonlinear representation methods that assumes high-dimensional obse...

Marginal Structural Model is a model for the marginal distribution or mean of counterfactual outcome...

Markdown is a lightweight markup language created by John Gruber in 2004 for formatting plain text d...

Markov Chain Monte Carlo is a family of algorithms that estimates expectations under a target distri...

Matérn Kernel is a stationary covariance function whose smoothness parameter controls the mean-squar...

Matthews Correlation Coefficient (MCC) is a correlation-based classification metric that uses all ce...

Maximum Mean Discrepancy (MMD) is an integral probability metric that measures the largest expectati...

MCP (Model Context Protocol) is an open, versioned protocol that lets AI Hosts use Clients to discov...

MCP Apps is an official optional Model Context Protocol extension that lets an MCP Server associate ...

MCP Client is the protocol component instantiated by an MCP Host to communicate directly with one MC...

MCP Elicitation is a Client capability that lets an MCP Server request additional user input during ...

MCP Extensions are optional, namespaced additions to the Model Context Protocol that define capabili...

MCP Gateway is an optional, MCP-aware proxy or aggregation layer that routes Client requests to Serv...

MCP Host is the user-facing AI application that creates and coordinates MCP Clients, assembles model...

MCP Prompt is a user-controlled, server-authored message template that an MCP Client can discover an...

MCP Resource is an application-controlled, URI-addressed data contract that an MCP Server exposes fo...

MCP Sampling is a deprecated Model Context Protocol Client feature that lets a Server request model ...

MCP Server is the protocol endpoint that exposes declared tools, resources, and prompts to an MCP Cl...

MCP Skills (Skills over MCP) is the official `io.modelcontextprotocol/skills` Extension for discover...

MCP Tasks is the official `io.modelcontextprotocol/tasks` Extension that lets an MCP Server return a...

MCP Tool is a model-controlled action contract that an MCP Server exposes for discovery and invocati...

MD5 (Message-Digest Algorithm 5) is a widely used cryptographic hash function that produces a 128-bi...

MDS (Multidimensional Scaling) is a family of methods that places objects in a low-dimensional coord...

Mechanistic Interpretability is the study of how a trained model implements a computation by identif...

A Membership Inference Attack is a privacy attack that uses a model's outputs, internal state, or tr...

Metadata Filtering is the enforcement of structured predicates over document or chunk attributes to ...

Metropolis-Hastings is a Markov Chain Monte Carlo algorithm that proposes a candidate state and acce...

Microsoft Agent Framework is Microsoft's open-source agent and workflow platform; its Python and .NE...

MIME Type is a standardized identifier used to indicate the nature and format of a file or data, con...

Minification is the process of removing all unnecessary characters from source code without changing...

Mixture of Experts (MoE) is a conditional-computation architecture in which a learned or fixed routi...

Mock data is data deliberately constructed for development, testing, or demonstration to represent a...

Model Alignment is the ongoing engineering and research work of making an AI system pursue intended ...

A Model Inversion Attack is a privacy attack that uses access to a trained model together with auxil...

Model Serving is the production system and operational practice that loads a versioned model artifac...

Monte Carlo Dropout is an uncertainty-estimation method that samples dropout masks during inference,...

MOTSA and sMOTSA are multi-object tracking and segmentation metrics that extend MOTA to pixel masks,...

Multi-Agent Systems are AI application architectures in which two or more bounded agents coordinate ...

A Multi-Armed Bandit is a sequential decision problem in which a learner repeatedly chooses one acti...

Multi-output Gaussian Process is a vector-valued Gaussian Process that jointly models several relate...

Multimodal AI is an AI system that accepts, relates, or generates more than one data modality, such ...

Multiple Hypothesis Tracking is a multi-target tracking framework that keeps several competing measu...

Multiple Kernel Learning (MKL) is a family of methods that jointly learns a predictor and the weight...

Multiple Object Tracking Accuracy is a CLEAR MOT score that subtracts missed detections, false detec...

Multiple Object Tracking Precision is a CLEAR MOT localization measure that averages the spatial sim...

Multiplicative Extended Kalman Filter is an attitude-estimation EKF that represents orientation with...

NanoID is a tiny, secure, URL-friendly unique string ID generator for JavaScript. It generates compa...

Network Dissection is a post-hoc interpretability method that quantifies how well an individual hidd...

Neural Network is a parameterized function represented as connected computational layers. Each layer...

NLP (Natural Language Processing) is the field that develops computational methods for analyzing, re...

No-U-Turn Sampler is an adaptive Hamiltonian Monte Carlo algorithm that builds a reversible trajecto...

Noise Sensitivity is a RAG evaluation metric that measures how often retrieved content induces incor...

Normalized Mutual Information is a symmetric partition-agreement measure that divides the mutual inf...

Nyström Method is a low-rank kernel approximation technique that selects landmark samples and recons...

OAuth (Open Authorization) is an open standard authorization protocol that allows users to grant thi...

Off-Policy Evaluation is the estimation of a target policy's expected value using logged outcomes co...

Ollama is an open-source model runtime and distribution tool that manages supported model artifacts ...

One-Class SVM is an unsupervised kernel method that estimates a boundary around a reference distribu...

Open-Weight Model is an AI model whose trained parameters are made available for download under stat...

OpenTelemetry is an open-source observability framework that provides a unified set of APIs, SDKs, a...

Optimal Sub-Pattern Assignment Metric is a normalized mathematical distance between finite point set...

ORPO (Odds Ratio Preference Optimization) is a direct preference optimization method that trains on ...

OSPA(2) Tracking Metric is a distance between finite sets of tracks that uses a time-averaged OSPA-d...

Out-of-Distribution Detection is the sample-level task of deciding whether an input is sufficiently ...

Overfitting is a failure of generalization in which a model or model-selection process adapts to fin...

PagedAttention is an LLM serving memory-management and attention approach that maps fixed-size logic...

Panoptic Segmentation is a computer vision task that assigns every image pixel a semantic class and ...

Panoptic Tracking (PAT) is an instance-centric panoptic tracking metric that combines frame-level Pa...

Parallel Tempering is a population Markov Chain Monte Carlo method that runs replicas at different t...

Part-Aware Panoptic Segmentation is a hierarchical image-understanding task that assigns every evalu...

Partial Dependence Plot (PDP) is a global model-inspection method that shows the average prediction ...

A Particle Filter is a Sequential Monte Carlo algorithm that approximates the filtering distribution...

Particle Markov Chain Monte Carlo is a family of methods that embeds particle filtering or condition...

PartPQ (Part-Aware Panoptic Quality) is a class-averaged metric that evaluates panoptic segment reco...

Password is a secret string of characters used to authenticate a user's identity and grant access to...

Patchscopes are configurable interpretability experiments that extract a hidden representation from ...

Path Patching is a mechanistic interpretability intervention that changes the contribution carried f...

PDF (Portable Document Format) is a file format developed by Adobe in 1992 to present documents cons...

PEFT (Parameter-Efficient Fine-Tuning) is a family of methods that adapts a pretrained model by trai...

Permutation Feature Importance (PFI) is a model-inspection method that measures the change in a fitt...

Planner-Executor is an AI agent architecture that creates a versioned, machine-readable plan for a g...

PNG (Portable Network Graphics) is a raster graphics file format that supports lossless data compres...

Poisson Multi-Bernoulli Mixture Filter is a Bayesian multi-object tracker that represents never-dete...

Policy Learning is the estimation of a rule that maps observed pre-treatment context to an action in...

Positivity Assumption is the causal-identification condition that every treatment being compared has...

A Posterior Predictive Distribution is the probability distribution of a new or replicated outcome a...

Potential Outcomes are the outcomes a unit would exhibit under each well-defined treatment or interv...

PPO (Proximal Policy Optimization) is a family of on-policy policy-gradient methods that improves a ...

PQ (Panoptic Quality) is a class-averaged metric for panoptic segmentation that combines the overlap...

Precision and Recall are complementary classification metrics: Precision is the share of predicted p...

Precision-Recall Curve is a threshold-sweep plot that places Recall on the horizontal axis and Preci...

Predictive Parity is a group fairness criterion requiring positive predictive value to be equal acro...

Preference Data is data that records judgments about candidate model behaviors as relative choices, ...

Prefill is the LLM inference phase that processes a request's uncached input tokens, produces the fi...

Prettify is the process of formatting code or data to make it more readable by adding proper indenta...

Principal Component Analysis (PCA) is a linear dimensionality-reduction method that projects centere...

Probabilistic Data Association Filter is a Bayesian tracking approximation that updates one known ta...

Probability Hypothesis Density Filter is a Random Finite Set filtering approximation that propagates...

Prompt is the complete input context supplied to a generative AI model for one inference request, co...

Prompt CI/CD is the controlled integration, evaluation, release, monitoring, and rollback of changes...

Prompt Engineering is the disciplined design, versioning, and evaluation of instructions, context, e...

Prompt Injection is attacker-controlled or attacker-influenced content that causes a large language ...

Prompt Regression Test is a repeatable evaluation that compares a candidate prompt or LLM applicatio...

Prompt Template is a versioned rendering contract that combines fixed instructions with validated ru...

Prompt Versioning is the practice of tracking prompt source history and binding each released LLM be...

Propensity Score is the conditional probability of receiving a treatment given observed pre-treatmen...

ProtoPNet is an interpretable image-classification architecture that compares latent patches from an...

Pseudo-Labeling is a training technique that treats selected predictions from a model or teacher as ...

Pseudo-Marginal MCMC is a Markov Chain Monte Carlo approach that replaces an unavailable likelihood ...

Qini Coefficient is an area-based metric that summarizes how far a model's cumulative incremental-ou...

QLoRA (Quantized Low-Rank Adaptation) is a parameter-efficient fine-tuning method that keeps a pretr...

QR Code (Quick Response Code) is a two-dimensional matrix barcode that can store various types of da...

Quantization is the process of mapping model weights, activations, or KV-cache values from a higher-...

Query Rewriting is a pre-retrieval transformation that converts a user's request and permitted conte...

Query String is the optional query component after the question mark (?) in a URL. Its exact syntax ...

R-Learner is a Causal Meta-Learner that estimates Conditional Average Treatment Effects by regressin...

RAG (Retrieval-Augmented Generation) is an architecture that retrieves external evidence at inferenc...

Random Finite Set is a set-valued random variable whose cardinality and member values are both rando...

Random Finite Set of Trajectories is a set-valued random variable whose elements are complete finite...

Random Fourier Features (RFF) are randomized explicit feature maps whose inner products approximate ...

Random Projection is a randomized linear dimensionality-reduction method that multiplies vectors by ...

Rao-Blackwellized Particle Filter is a Sequential Monte Carlo method that samples one subset of late...

Rate limiting is a technique used to control the number of requests a client can make to an API or s...

A Rauch-Tung-Striebel Smoother is a fixed-interval Bayesian algorithm that runs a Kalman Filter forw...

RBF Kernel (Radial Basis Function Kernel) is a positive-definite kernel that assigns similarity `exp...

Reciprocal Rank Fusion is an unsupervised rank-aggregation method that scores each document by summi...

Recommendation System is a decision system that selects and orders a limited set of items for a user...

Red Teaming is a structured, authorized adversarial assessment that attempts to cause harmful, insec...

Refusal Correctness is an evaluation property that measures whether a RAG system answers answerable ...

Regex (Regular Expression) is a sequence of characters that defines a search pattern, primarily used...

Regression Discontinuity Design is a quasi-experimental design that identifies a local causal effect...

Regret is a performance measure for sequential decisions that compares the reward or loss accumulate...

Reinforcement Learning (RL) is a machine-learning paradigm in which an agent learns a policy for seq...

Reproducing Kernel Hilbert Space (RKHS) is a Hilbert space of functions in which evaluating a functi...

Reranking is a second-stage ranking operation that scores a query together with already retrieved ca...

REST (Representational State Transfer) is an architectural style for designing networked application...

Retriever is a query-to-evidence component that applies a declared search contract and returns ranke...

Reversible Jump MCMC is a trans-dimensional Markov Chain Monte Carlo method that samples a model ind...

Reward Model is a learned model that maps a context and candidate behavior to a scalar or structured...

RGB Color (Red Green Blue) is an additive color model where red, green, and blue light are combined ...

RGBA Color is an extension of the RGB color model that adds an alpha channel for opacity control, al...

RLHF (Reinforcement Learning from Human Feedback) is a family of training methods that use human jud...

RNN (Recurrent Neural Network) is a neural-network architecture that processes an ordered sequence b...

Robust Aggregation is a family of rules for combining gradients or model updates while limiting the ...

ROC-AUC is the area under the Receiver Operating Characteristic curve, summarizing how often a scori...

Router Agent is an AI-assisted or deterministic routing component that maps a request to one or more...

RSA (Rivest-Shamir-Adleman) is an asymmetric cryptographic algorithm that uses a pair of keys - a pu...

Secure Aggregation is a cryptographic protocol that lets a coordinator compute an aggregate, typical...

Selective Prediction is a decision framework in which a predictor is paired with a selection functio...

Self-Supervised Learning is a learning paradigm that constructs supervision from the structure, cont...

Semantic Cache is a cache that embeds an incoming request, finds a sufficiently similar cached reque...

Semantic Search is information retrieval that uses learned representations to find documents related...

Semi-Supervised Learning is a machine learning setting in which an algorithm uses both labeled examp...

Sentiment Analysis is an NLP task that assigns a declared sentiment representation, such as polarity...

Sequential Monte Carlo is a family of algorithms that approximates a sequence of probability distrib...

Serialization is the process of converting complex data structures or objects into a format that can...

SFT (Supervised Fine-Tuning) is a fine-tuning method that continues training a pretrained model on l...

SHA-256 (Secure Hash Algorithm 256-bit) is a cryptographic hash function that produces a 256-bit (32...

SHAP (SHapley Additive exPlanations) is a framework for explaining a model prediction with additive ...

Silhouette Score is an internal cluster-validity measure that compares each sample's average distanc...

Skill, in the Agent Skills sense, is a reusable directory centered on a SKILL.md file that gives a c...

Slice Sampling is a Markov Chain Monte Carlo method that augments an unnormalized target density wit...

Slug is a URL-friendly version of a string, typically derived from a title or name, that uses only l...

Small Language Model (SLM) is an informal, context-dependent category for a language model that has ...

Sparse Autoencoder is an encoder-decoder model that reconstructs an input through a latent represent...

Sparse Gaussian Process is a family of approximations that represents a Gaussian Process through a s...

Sparse Retrieval is a first-stage search family that represents a query and document with high-dimen...

Specificity is the proportion of actual negative examples that a classifier correctly predicts as ne...

Spectral Clustering is a graph-based method that embeds samples with eigenvectors of a similarity-gr...

Speculative Decoding is an autoregressive inference method that proposes multiple future Tokens with...

Speech Recognition, also called automatic speech recognition (ASR) or speech-to-text (STT), converts...

sPTQ is a class-averaged panoptic tracking metric that subtracts the IoU credit of matched segments ...

SQL (Structured Query Language) is a standardized programming language designed for managing and man...

A Square-Root Kalman Filter is a Kalman Filter implementation that propagates a matrix factor of sta...

SSE Transport in MCP usually means the legacy HTTP+SSE transport that opened a persistent GET event ...

SSL/TLS (Secure Sockets Layer / Transport Layer Security) are cryptographic protocols designed to pr...

State Space Model (SSM) represents a dynamic system or sequence with a latent state that is updated ...

STDIO Transport is a Model Context Protocol transport in which an MCP client launches a server subpr...

STQ (Segmentation and Tracking Quality) is a video panoptic segmentation metric that combines semant...

Structural Causal Model (SCM) is a formal system of autonomous structural assignments that maps exog...

Structured Output is an LLM inference capability that makes a response conform to a declared machine...

Superposition is a neural representation strategy in which a model encodes more features than the av...

Supervised Learning is a machine learning paradigm that fits a function from inputs to declared targ...

Supervisor Agent is the centralized control-loop role in a multi-agent system that owns shared task ...

Support Vector Machine (SVM) is a supervised learning method that selects a decision boundary with a...

Support Vector Regression (SVR) is a supervised regression method that seeks a regularized function ...

SVG (Scalable Vector Graphics) is an XML-based vector image format for two-dimensional graphics that...

Synthetic Control Method is a comparative-case design that estimates a treated unit's counterfactual...

Synthetic Data is data produced by rules, simulations, statistical models, or generative models rath...

System Prompt is a provider-specific high-authority instruction layer used to define persistent beha...

t-SNE (t-Distributed Stochastic Neighbor Embedding) is a nonlinear visualization method that places ...

Target Trial Emulation is a design framework that specifies the protocol of a hypothetical randomize...

TCAV (Testing with Concept Activation Vectors) is a post-hoc interpretability method that learns a d...

Temperature is a decoding parameter that divides next-token logits before softmax, sharpening the pr...

Tensor Parallelism is an intra-layer model-parallel strategy that shards a tensor operation and its ...

Test-Time Compute is the explicit allocation of additional inference computation to a problem throug...

Text-to-Image is a generative AI task that maps a natural-language prompt, and sometimes additional ...

Text-to-Video is a generative AI task that maps a natural-language prompt, and sometimes additional ...

Thompson Sampling is a randomized sequential decision policy that samples one plausible model from t...

Throughput is the amount of explicitly defined work an LLM serving system completes per unit of wall...

TID and LGD are nuScenes tracking diagnostics that measure how long a ground-truth trajectory waits ...

Time-Varying Confounding is confounding in longitudinal data where a variable predicts later treatme...

A timestamp is a sequence of characters or encoded information identifying when a certain event occu...

Timezone is a region of the globe that observes a uniform standard time for legal, commercial, and s...

Token is a model-specific unit represented by an integer ID that a language model consumes or produc...

Tokenizer is the model-bound pipeline that transforms text or structured messages into token IDs and...

Tokens per Second is a rate that must be qualified by token type, scope and timing interval, such as...

TOML is a configuration file format designed to be easy to read due to its obvious semantics, mappin...

Tool Poisoning is an AI-agent attack in which malicious or compromised tool metadata, schemas, imple...

Tool Use is the system pattern in which a model selects or proposes an external capability, trusted ...

Track Completeness and Fragmentation are trajectory diagnostics that classify each ground-truth iden...

Track Mean Average Precision is a confidence-ranked tracking metric that computes average precision ...

Tracking Every Thing Accuracy is a large-vocabulary multi-object tracking metric that separately mea...

Training Data is the versioned collection of examples, signals, and associated metadata that a learn...

Trajectory Cardinality Probability Hypothesis Density Filter is an assumed-density multi-object trac...

Trajectory Generalized Optimal Sub-Pattern Assignment Metric is a mathematical distance between fini...

Trajectory Poisson Multi-Bernoulli Mixture Filter is a Bayesian multi-object tracker that represents...

Trajectory Probability Hypothesis Density Filter is an assumed-density multi-object tracker that pro...

Transformer is a neural-network architecture that repeatedly updates a sequence of vectors with atte...

Truncated SVD (Truncated Singular Value Decomposition) is a matrix-factorization method that compute...

TTFT (Time to First Token) is the client-observed elapsed time from sending a streaming LLM request ...

The Tuned Lens is an interpretability method that trains a separate affine translator for each model...

TypeScript is a strongly typed programming language that builds on JavaScript by adding optional sta...

UMAP (Uniform Manifold Approximation and Projection) is a nonlinear dimensionality-reduction method ...

Uncertainty Quantification is the discipline of defining, estimating, validating, and using uncertai...

Unicode is a universal character encoding standard that assigns a unique number (code point) to ever...

Unit conversion is the process of expressing the same physical or measured quantity in a different u...

Unix Timestamp (also known as Unix Epoch Time or POSIX time) is a system for tracking time as a runn...

An Unscented Kalman Filter is a nonlinear Gaussian state estimator that propagates deterministically...

Unsupervised Learning is a machine learning paradigm that optimizes a structure-seeking objective on...

Uplift Modeling is the use of experimental or causally identified data to estimate or rank how an in...

Upper Confidence Bound is an optimism-based action-selection principle that chooses the option with ...

URI is a string of characters that identifies a resource on the internet or within a system, providi...

URL is a specific type of URI that provides the complete address to locate a resource on the interne...

URL Encoding (Percent-Encoding) is a mechanism for encoding information in a Uniform Resource Identi...

UTF-8 (8-bit Unicode Transformation Format) is a variable-width character encoding that can represen...

UUID (Universally Unique Identifier) is a 128-bit identifier format standardized by RFC 9562. The ca...

VAE (Variational Autoencoder) is a latent-variable generative model that trains an encoder to approx...

Variational Inference is a family of methods that approximates an intractable target distribution wi...

A vector database is a datastore that persists vector embeddings and retrieves nearby records under ...

Video Panoptic Segmentation is a dense video-understanding task that assigns every evaluated pixel a...

vLLM is an open-source library and server runtime for executing supported language, multimodal, embe...

VLM (Vision-Language Model) is a model trained to connect visual and linguistic representations so i...

VPQ (Video Panoptic Quality) is a windowed video panoptic segmentation metric that applies Panoptic ...

Weak Supervision is a family of learning settings in which the available training signal is incomple...

WebLLM is an open-source, in-browser language-model inference engine from MLC AI. It loads compatibl...

WebP is a modern image format developed by Google that provides superior lossless and lossy compress...

Whitespace refers to characters that represent horizontal or vertical space in text but are typicall...

Workflow Orchestration is the structured coordination of tasks, dependencies, state, timers, signals...

X-Learner is a Causal Meta-Learner that estimates response surfaces, imputes treatment effects separ...

XML (Extensible Markup Language) is a markup language and file format for storing, transmitting, and...

XPath is a query language for selecting nodes and computing values from XML documents using path exp...

YAML (YAML Ain't Markup Language) is a human-readable data serialization language commonly used for ...

Zero-Shot Learning is a family of transfer and evaluation settings in which a model predicts target ...