AI Hub

Glossary

A plain-English glossary of the methods and ideas behind modern AI. Search, sort, or jump to a letter.

98 terms

Definition
AblationAn experiment that removes or changes one component of a model or training setup to measure how much it actually contributes.May 22, 2026
Adam OptimizerA widely-used adaptive optimizer for training neural networks.May 21, 2026
Agentic Reinforcement LearningTraining LLMs with RL where the model takes multi-step actions in an environment (tools, code, web search) and is rewarded on task outcomes.May 21, 2026
AI AgentsSystems where a model plans, uses tools, and acts over multiple steps.May 21, 2026
AI AlignmentMaking AI systems pursue what people actually intend.May 21, 2026
AttentionA mechanism that lets a model weigh which other tokens matter for each token.May 21, 2026
AutoencoderA network that compresses data and reconstructs it.May 21, 2026
Autoregressive ModelsModels that generate output one token at a time, left to right.May 21, 2026
BackpropagationThe algorithm that trains neural networks by propagating error gradients.May 21, 2026
Batch NormalizationNormalizing layer activations to stabilize and speed up training.May 21, 2026
Chain-of-Thought PromptingPrompting a model to reason step by step before answering.May 21, 2026
CLIPA model that learns shared text and image embeddings from paired data.May 21, 2026
Constitutional AIAligning models using a written set of principles instead of human labels.May 21, 2026
Context EngineeringSystematically constructing and evolving an agent’s context — instructions, memory, retrieved info, tool results — as the primary lever for performance.May 21, 2026
Context WindowThe maximum number of tokens a model can attend to at once, spanning both the prompt and its generated output.May 22, 2026
Contrastive LearningLearning by pulling related items together and pushing others apart.May 21, 2026
Convolutional Neural NetworkA network that uses sliding filters, ideal for images.May 21, 2026
Data ContaminationWhen benchmark or test data leaks into a model’s training set, inflating its scores.May 22, 2026
Deep LearningMachine learning with many-layered neural networks that learn features automatically.May 21, 2026
Deep Research AgentsAutonomous agents that iteratively search, read, reason, and synthesize multi-source, cited reports on a question over many steps.May 21, 2026
DeepSeek Sparse Attention (DSA)A fine-grained trainable sparse attention built on Multi-head Latent Attention, using a lightning indexer plus token selection.May 21, 2026
Diffusion ModelsGenerative models that create data by gradually denoising random noise.May 21, 2026
Direct Preference OptimizationA simpler alternative to RLHF that optimizes preferences without a reward model.May 21, 2026
DropoutRandomly disabling neurons during training to prevent overfitting.May 21, 2026
EmbeddingsNumeric vectors that capture the meaning of text, images, or other data.May 21, 2026
Emergent AbilitiesCapabilities that appear only once models get large enough.May 21, 2026
EpochOne full pass over the entire training dataset.May 22, 2026
Expert SystemsClassic AI based on hand-coded rules and knowledge.May 21, 2026
Few-Shot LearningPerforming a task from just a handful of examples.May 21, 2026
Fine-tuningAdapting a pretrained model to a specific task or domain with further training.May 21, 2026
FlashAttentionAn exact, IO-aware attention algorithm that is much faster and uses less memory.May 21, 2026
FLOPsFloating-point operations — the standard unit for measuring the compute used to train or run a model.May 22, 2026
Foundation ModelsLarge models pretrained broadly, then adapted to many tasks.May 21, 2026
Generative Adversarial NetworkTwo networks — a generator and a critic — trained against each other.May 21, 2026
Gradient DescentIteratively nudging parameters downhill to minimize a loss.May 21, 2026
Group Relative Policy Optimization (GRPO)A critic-free RL algorithm that estimates the advantage baseline from a group of sampled responses to the same prompt.May 21, 2026
Group Sequence Policy Optimization (GSPO)A reasoning-RL algorithm that computes importance ratios and clips at the whole-sequence level rather than per token.May 21, 2026
Grouped-Query AttentionAn attention variant that shares key/value heads to speed inference.May 21, 2026
HallucinationWhen a model states false information confidently.May 21, 2026
HyperparameterA configuration value set before training (e.g. learning rate, batch size) rather than learned by the model.May 22, 2026
In-Context LearningLearning a task from examples in the prompt, without weight updates.May 21, 2026
InferenceRunning a trained model to produce outputs — as opposed to training, which updates its weights.May 22, 2026
Instruction TuningFine-tuning a base model to follow natural-language instructions.May 21, 2026
Interactive World ModelsGenerative models that produce explorable, action-conditioned 3D/video environments in real time from a prompt.May 21, 2026
Knowledge DistillationTraining a small "student" model to imitate a larger "teacher".May 21, 2026
KV CacheCached attention keys and values that let a model generate each new token without recomputing the whole sequence.May 22, 2026
Latent ReasoningPerforming reasoning in the model’s continuous hidden-state space instead of by emitting explicit chain-of-thought tokens.May 21, 2026
Layer NormalizationNormalizing across features within each token.May 21, 2026
LogitsA model’s raw, unnormalized output scores for each possible next token, converted to probabilities by softmax.May 22, 2026
LoRALow-Rank Adaptation — efficient fine-tuning by training small adapter matrices.May 21, 2026
LSTMA recurrent network with gates that remember long-range information.May 21, 2026
Masked Language ModelingPretraining by predicting hidden (masked) tokens.May 21, 2026
Mixture of ExpertsA sparse architecture that routes each token to a few specialized sub-networks.May 21, 2026
Mixture-of-Recursions (MoR)An architecture that recursively reuses shared layers and routes each token to its own recursion depth for adaptive computation.May 21, 2026
Model Context Protocol (MCP)An open standard (JSON-RPC) for connecting LLMs and agents to external tools, data, and resources via a uniform client–server interface.May 21, 2026
Monte Carlo Tree SearchA search method that samples and evaluates promising move sequences.May 21, 2026
Multimodal LearningModels that jointly understand multiple data types.May 21, 2026
Native Sparse Attention (NSA)A hardware-aligned, end-to-end-trainable sparse attention using hierarchical token compression and selection.May 21, 2026
Neural NetworkA model of interconnected "neurons" that learns patterns from data.May 21, 2026
OverfittingWhen a model memorizes its training data and fails to generalize to new, unseen inputs.May 22, 2026
ParametersThe learned weights of a neural network; the "parameter count" (e.g. 70B) is a rough proxy for a model’s size and capacity.May 22, 2026
PerceptronThe earliest trainable artificial neuron.May 21, 2026
PerplexityA measure of how well a language model predicts a sample of text; lower is better.May 22, 2026
Positional EncodingHow Transformers represent the order of tokens.May 21, 2026
PretrainingThe large-scale self-supervised stage that builds a model’s base knowledge.May 21, 2026
Prompt EngineeringCrafting inputs to steer a model toward better outputs.May 21, 2026
QuantizationShrinking models by storing weights at lower numerical precision.May 21, 2026
Reasoning ModelsModels that spend extra compute "thinking" before they answer.May 21, 2026
Recurrent Neural NetworkA network with memory, processing sequences one step at a time.May 21, 2026
Reinforcement LearningLearning by trial and error to maximize reward.May 21, 2026
Residual ConnectionsSkip connections that let very deep networks train.May 21, 2026
Retrieval-Augmented GenerationGrounding model answers in documents fetched at query time.May 21, 2026
Reward ModelA model trained to score outputs by human preference, used to guide reinforcement learning during alignment.May 22, 2026
RLHFReinforcement Learning from Human Feedback — aligning models to human preferences.May 21, 2026
Rotary Position EmbeddingA positional scheme that rotates query/key vectors by position.May 21, 2026
Scaling LawsEmpirical relationships predicting model quality from compute, data, and size.May 21, 2026
Self-PlayImproving by training against copies of oneself.May 21, 2026
Self-Supervised LearningLearning from unlabeled data by predicting part of it from the rest.May 21, 2026
Sequence-to-SequenceAn encoder-decoder framework that maps one sequence to another.May 21, 2026
Speculative DecodingSpeeding up generation with a small draft model the big one verifies.May 21, 2026
State Space ModelsSequence models with linear-time scaling, an alternative to attention.May 21, 2026
Supervised Fine-Tuning (SFT)Training a pretrained model on labeled input→output examples to teach a specific behavior, format, or task.May 22, 2026
Supervised LearningLearning from labeled input-output examples.May 21, 2026
Synthetic DataTraining data generated by models rather than collected from humans.May 22, 2026
TemperatureA sampling setting that controls randomness — low values make output focused and deterministic, high values make it more diverse.May 22, 2026
Test-Time Compute ScalingImproving accuracy by spending more computation at inference — longer reasoning or multiple samples — rather than only at training.May 21, 2026
TokenThe atomic unit of text a model reads and generates — typically a word, sub-word, or character chunk.May 22, 2026
TokenizationSplitting text into the discrete units a model actually reads.May 21, 2026
Tool UseLetting models call external tools, code, and APIs.May 21, 2026
Top-p (Nucleus) SamplingSampling the next token from the smallest set of candidates whose probabilities sum to p.May 22, 2026
Transfer LearningReusing a model trained on one task to bootstrap another.May 21, 2026
TransformerThe neural-network architecture behind virtually all modern language models.May 21, 2026
Tree of ThoughtsReasoning by exploring and evaluating multiple solution paths.May 21, 2026
Turing TestAlan Turing’s thought experiment for machine intelligence.May 21, 2026
Unsupervised LearningFinding structure in data without labels.May 21, 2026
Variational AutoencoderAn autoencoder that learns a smooth, samplable latent space.May 21, 2026
Vision TransformerApplying the Transformer architecture directly to images.May 21, 2026
Zero-Shot LearningDoing a task with no examples, from instructions alone.May 21, 2026