Abstract
Hierarchical Federated Learning (HFL) requires a scalable and transparent incentive mechanism, yet existing on-chain approaches are too costly and centralized solutions lack trust. To address this, we propose CoFT, a hybrid framework that manages rewards off-chain using a hierarchy of state channels and a stablecoin while securing the system’s integrity with a single on-chain cryptographic anchor. This architecture dramatically reduces on-chain complexity from a prohibitive O(N×T) to a sustainable O(I+S+N), enabling cryptographic verifiability, automated trustless payouts, and fair burden-sharing. CoFT thus offers a robust and economically viable blueprint for the financial infrastructure of large-scale, trustworthy collaborative AI ecosystems.
Funding Information
  • INNOPOLIS Foundation (1711202494)
  • IITP (Institute of Information & Communications Technology Planning & Evaluation)-ITRC (IITP-2025-RS-2020-II201797)

This publication has 26 references indexed in Scilit: