Activation Functions & Normalization Variants
7 definitions, including Batch normalization, GELU (Gaussian Error Linear Unit), Group normalization.
Glossary
Inference terms, in plain English.
Additional Architecture & Training Detail
12 definitions, including Activation steering, Attention head dimension, Hidden size / model dimension.
Additional Core Benchmarks & Evaluation Concepts
6 definitions, including BEIR, Cost-quality frontier, Long-context degradation curve.
Additional Evaluation Benchmarks & Domains
14 definitions, including Aider polyglot benchmark, Codeforces rating (LLM), FinanceBench.
Additional Long-Context & Retrieval Techniques
8 definitions, including Infini-attention, Landmark attention, LongRoPE.
Additional Model Concepts & Prompting Techniques
8 definitions, including Chain-of-verification, In-context learning, Meta-prompting.
Additional Quantization & Compression Detail
11 definitions, including bitsandbytes, Blockwise quantization, Double quantization.
Additional Safety, Evaluation & Model Behavior Terms
10 definitions, including Distillation restriction clause, Gated/restricted model release, Model lineage.
Additional Sampling, Decoding & Reasoning Techniques
9 definitions, including Acceptance rate (speculative decoding), Diverse beam search, Draft tree / tree-based speculative decoding.
Additional Serving Reliability & Observability Terms
8 definitions, including Circuit breaker (inference routing), Health check (model endpoint), Jitter (latency).
Additional Terms Rounding Out Coverage
10 definitions, including Cascade routing, Compositional generalization, Confidence-based routing.
Additional Tokenizer & Language Coverage Terms
5 definitions, including Context length in characters vs tokens, Fertility (tokenizer), Merge rules (BPE).
Agents & harnesses
102 definitions, including agent framework, agent harness, agent identity and permissions.
Agents & Tool Use
8 definitions, including Agentic loop, Computer use / GUI agent, Function calling / tool calling.
Attention Mechanisms & Transformer Architecture
22 definitions, including Attention sink, Causal attention mask, Cross-attention.
Classic Architectures (for contrast)
5 definitions, including Convolutional neural network (CNN), Gated recurrent unit (GRU), Long short-term memory (LSTM).
Classic Optimizers & Training Infrastructure Detail
8 definitions, including Adam optimizer, Gradient accumulation, Gradient clipping.
Cost & FinOps
141 definitions, including AI API cost estimator tool, AI API pricing comparison, AI budget justification.
Data Engineering for Model Training
10 definitions, including Common Crawl, Data contamination detection, Data mixture / mixture weights.
Deployment & security
110 definitions, including A/B testing AI models, agentic AI governance, AI agent orchestration.
Distributed Training & Hardware Systems
11 definitions, including Accelerator interconnect topology, All-gather, All-reduce.
Ecosystem & products
65 definitions, including AGI (artificial general intelligence), AI app builder, AI engineer.
Embeddings & Representations
14 definitions, including Approximate nearest neighbor (ANN) search, BM25, Chunking.
Evals & quality
56 definitions, including adversarial testing (LLM), AgentBench, answer relevance.
Evaluation Benchmarks
25 definitions, including AlpacaEval, ARC (AI2 Reasoning Challenge), Arena-Hard.
Fine-Grained Serving & Inference Optimization Terms
13 definitions, including Batch size (inference), Canary deployment (model), Continuous integration for model evaluation.
Fine-Tuning & Alignment Methods
7 definitions, including Adapter layers, DPO (Direct Preference Optimization), GRPO (Group Relative Policy Optimization).
Governance & compliance
131 definitions, including AI acceptable use policy, AI accountability framework, AI Act penalties.
Hardware
47 definitions, including A100, B200, bare-metal GPU.
Inference & serving
124 definitions, including acceptance rate, admission control, arithmetic intensity.
Inference API & Deployment Concepts
9 definitions, including Fallback routing, Idempotency key, Load shedding.
Information Theory, Statistics & Core Math
18 definitions, including Backpropagation, Bias-variance tradeoff, Bootstrap resampling.
Model Architecture Families & Training Concepts
21 definitions, including AdamW optimizer, Autoregressive model, Batch size (training).
Model Evaluation Metrics & Statistical Concepts
7 definitions, including Accuracy, BERTScore, BLEU score.
Multi-Model & Merging Techniques
4 definitions, including DARE (Drop And REscale), Frankenmerge / passthrough merge, Mixture of agents.
Multimodal & Vision-Language Terminology
14 definitions, including Automatic speech recognition (ASR), Autoregressive image generation, CLIP (Contrastive Language-Image Pretraining).
Positional Encoding & Context Length
9 definitions, including Absolute positional encoding, ALiBi (Attention with Linear Biases), Context length extrapolation.
Privacy & governance
35 definitions, including AI bill of materials (AIBOM), AI risk management, AI usage policy.
Procurement & vendor
71 definitions, including AI budget approval process, AI contract negotiation, AI integration partner.
Quantization
18 definitions, including AWQ (Activation-aware Weight Quantization), BF16 (bfloat16), Calibration dataset.
Routing & cost
73 definitions, including A/B testing (models), AI gateway, blended cost per token.
Safety & Alignment Terminology
10 definitions, including Alignment tax, Backdoor attack (model), Differential privacy.
Sampling, Decoding & Generation Parameters
18 definitions, including Contrastive search, Frequency penalty, Greedy decoding.
Serving, Batching & Systems
10 definitions, including Autoscaling, EAGLE, Expert parallelism (MoE serving).
Speculative Decoding & Model Compression (Distillation, Pruning)
6 definitions, including Layer skipping / early exit, Model pruning, Structured pruning.
Strategy & adoption
71 definitions, including agentic AI adoption roadmap, agentic AI enterprise adoption, AI adoption barriers enterprise.
Tokenization
9 definitions, including Byte-level BPE, Detokenization, Out-of-vocabulary (OOV) token.
Training & tuning
97 definitions, including active parameters, adapters (fine-tuning), agentic RL.
uncategorized
1 definitions, including Format: term.
Use cases
74 definitions, including AI accessibility tools enterprise, AI agent for procurement, AI chatbot for customer service.