Correlation-Aware Liquidation Sequencing, Cross-Chain Dependence Synchronisation, Off-Chain Dependence Bootstrapping, and Continuous Regime-Adaptive Risk Control for On-Chain Lending Pools
Year: 2026
Type: article
Abstract: On-chain lending protocols that price collateral by measured pairwise dependence between assets need more than a health function. This disclosure describes four mechanisms that extend such a protocol to multi-asset portfolios, multiple blockchains and newly tokenised assets. (1) Liquidation sequencing scores each collateral asset by the reduction in the portfolio's value-weighted average dependence that seizing it would produce. Assets are liquidated in order of decreasing score, with the scores recomputed after each partial liquidation. (2) A cross-chain dependence oracle exchanges timestamped price observations between chains. It computes dependence scores for assets on different chains and injects them into each chain's local oracle under asymmetric rate limiting: decreases are bounded, and increases apply immediately. (3) Off-chain bootstrapping initialises dependence scores for newly tokenised assets from signed attestations of historical market data. A trust discount shifts the imported score toward maximum correlation, and a mandatory decay can only tighten it. On-chain observations replace it as they accumulate. (4) Continuous regime-adaptive control computes a weighted dependence indicator over all asset pairs. Monotonic transfer functions with asymmetric smoothing map it to risk parameters, in place of a binary circuit breaker. A constant-time maintenance scheme for the indicator is given. Variants, alternative embodiments and sixteen dependent claims are included. The mechanisms interoperate with an asymmetric health function, a rate-limited oracle, a deposit guard, and stock and flow liquidation bounds described in a companion document.
Author: Zisckind, Tal
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7224814287