An educational crypto, finance, and AI vocabulary puzzle about rollup posting, time-value math, model decoding, and dataset cleanup.
Rollup Mechanics
These terms cover bundled submissions, execution data, dispute periods, and invalid claims.
Batch PostingCalldataChallenge WindowFraud Proof
- Batch Posting: Batch posting groups many transactions or messages into one rollup submission.
- Calldata: Calldata is transaction data carried on chain, often used to publish rollup execution data.
- Challenge Window: The challenge window is the time when someone can contest a proposed rollup state.
- Fraud Proof: A fraud proof shows that a claimed state transition or execution result is invalid.
Yield Math
These terms cover interest buildup, reinvestment, discounting, and loan repayment schedules.
AccrualCompoundingDiscountingAmortization
- Accrual: Accrual is the process of interest or income accumulating over time.
- Compounding: Compounding means earnings are added back so future growth is calculated on a larger base.
- Discounting: Discounting converts a future cash amount into its present value.
- Amortization: Amortization spreads a loan or asset cost across a series of time periods.
Decoding Controls
These terms cover raw scores, probabilistic choice, deterministic choice, and nucleus sampling.
LogitsSamplingGreedy DecodingTop-p
- Logits: Logits are the raw scores a model produces before they are normalized into probabilities.
- Sampling: Sampling selects the next token using probabilities instead of always taking the top choice.
- Greedy Decoding: Greedy decoding always chooses the highest-probability next token at each step.
- Top-p: Top-p sampling draws the next token from the smallest set whose combined probability passes a chosen cutoff.
Data Hygiene
These terms cover labels, descriptive fields, duplicates, and rule checks.
AnnotationMetadataDeduplicationValidation
- Annotation: Annotation is a human-added label, note, or tag attached to a piece of data.
- Metadata: Metadata is descriptive information that helps explain or organize a record or dataset.
- Deduplication: Deduplication is the process of removing duplicate items from a dataset.
- Validation: Validation checks that data matches expected rules, formats, or constraints.