Aug 20, 2026

On-Chain Credit, Price Discovery, Retrieval Stack, and Evaluation Oversight

An educational crypto, finance, and AI vocabulary puzzle about collateral rules, market quotes, retrieval systems, and model review.

On-Chain Credit Controls

These terms cover collateral, risk buffers, forced liquidation, and credit limits.

  • Collateral Ratio: The collateral ratio compares the value posted as collateral with the value of the debt or exposure it backs.
  • Health Factor: The health factor is a risk metric that shows how close a borrowing position is to liquidation.
  • Liquidation Threshold: The liquidation threshold is the collateral level at which a position can be forcibly closed.
  • Borrow Cap: A borrow cap is the maximum amount that can be borrowed against a market, asset, or protocol reserve.

Price Discovery Basics

These terms cover market depth, quoted gaps, execution drag, and size increments.

DepthSpreadSlippageTick Size
  • Depth: Depth is the amount of buy and sell interest available around the current market price.
  • Spread: The spread is the difference between the best bid and the best ask in a market.
  • Slippage: Slippage is the difference between the expected price and the price a trade actually gets.
  • Tick Size: Tick size is the minimum price step a market allows between quoted levels.

Retrieval Stack

These terms cover vectors, search, ranking, and the store that supports them.

EmbeddingRetrievalRerankerVector Database
  • Embedding: An embedding is a numeric representation that captures meaning or similarity.
  • Retrieval: Retrieval is the act of fetching useful external information for a model or agent.
  • Reranker: A reranker is a model that reorders retrieved items to put the most relevant ones first.
  • Vector Database: A vector database stores embeddings so similar items can be searched efficiently.

Evaluation Oversight

These terms cover harnesses, adversarial review, isolation, and human checkpoints.

Eval SuiteRed TeamingSandboxingHuman in the Loop
  • Eval Suite: An eval suite is a set of tests and metrics used to measure how a model or agent behaves.
  • Red Teaming: Red teaming is the practice of probing a system for weaknesses by trying to make it fail.
  • Sandboxing: Sandboxing runs code or actions in a restricted environment with limited permissions.
  • Human in the Loop: Human in the loop means a person reviews, corrects, or approves key outputs or actions.