Detecting Inconsistencies Between DAO Proposals and Codes

Level: MA/MAP
Responsible person: Guangyao Li
Keywords:  DAO, Blockchain, Tokenomics, Token Engineering

Decentralized Autonomous Organizations (DAOs) rely on transparent and automated governance through blockchain-based smart contracts. Typically, one DAO proposal is written in natural language, but the on-chain execution logic is written in smart contracts. However, a critical issue arises: the text of a proposal may not always match the actual code changes that are executed on-chain, which might be caused by human error, ambiguous or misleading descriptions, or Malicious activities. Such inconsistencies can lead to security risks, governance failures, and loss of community trust. 

This project/thesis aims to develop frameworks using LLMs to analyze and compare DAO proposals with their corresponding on-chain code to detect potential inconsistencies or mismatches. Specific objectives include: 

1. Collect governance proposals and related smart contract code from DAOs (e.g., Aragon, Snapshot, Compound, Aave). 

2. Extract and represent the intended actions described in the proposal text. 

3. Analyze the actual logic or parameters implemented in the smart contract execution. 

4. Develop methods to detect inconsistencies between the proposal’s description and the resulting code behavior. In the end, provide visualization or reporting tools to highlight detected mismatches.