An educational crypto, finance, and AI vocabulary puzzle about backing, order books, training, and inference flow.
Stablecoin Mechanics
These terms cover reserve levels, verification, redemption rights, and the range around a target peg.
Reserve RatioAttestationRedemption RightPeg Band
- Reserve Ratio: The reserve ratio compares backing assets with the amount of stablecoin supply they support.
- Attestation: An attestation is a statement or report that verifies reserves or other claims.
- Redemption Right: A redemption right is the ability to swap tokens back for the underlying asset or value.
- Peg Band: A peg band is the range around a target price in which a stablecoin is considered near parity.
Market Microstructure
These terms focus on quotes, available size, and the price movement caused by large orders.
- Order Book: An order book shows outstanding buy and sell orders, usually sorted by price.
- Market Depth: Market depth measures how much volume is available at nearby prices.
- Bid-Ask Spread: The bid-ask spread is the difference between the highest bid and the lowest ask.
- Price Impact: Price impact is the amount a trade moves the market while it is being executed.
Model Training
These terms describe how text is split, how models are adapted, and how larger systems are compressed.
TokenizerFine-TuningDistillationMixture of Experts
- Tokenizer: A tokenizer converts text into tokens that an AI model can process.
- Fine-Tuning: Fine-tuning updates a pretrained model on a narrower dataset or task.
- Distillation: Distillation trains a smaller model to imitate the behavior of a larger one.
- Mixture of Experts: A mixture of experts uses routing to send input to specialized submodels.
Serving Pipelines
These terms describe the steps from prompt processing to token streaming and traffic management.
PrefillToken StreamingRequest RoutingAutoscaling
- Prefill: Prefill is the initial pass that processes the input prompt before generation begins.
- Token Streaming: Token streaming delivers model output incrementally instead of waiting for the full answer.
- Request Routing: Request routing chooses which service or worker handles a given model request.
- Autoscaling: Autoscaling changes available capacity based on demand so a system can keep up with traffic.