Validation
Graph 1: Steps to Data Validation
The data submitted by the user will pass our 4-stage validation, as shown in Figure 1:
Before AI Validation: This stage mainly verifies several basic rules of data, such as completion of required fields and address alignment.
AI Validation: This stage relies on multi-modal LLM capability support, powered by Claude, Gemini, and OpenAI. Multi-modal validation is reflected in the validity of transaction addresses, tag categories, transaction hashes, text descriptions, evidence screenshots in the provided evidence data, as well as the validity of cross comparisons.
Human Intelligence: This stage is based on manual verification and is divided into two rounds, namely public voting and domain expert voting. Platform users participate in public voting, and domain experts from the platform and its partners participate in domain expert voting. The voting results will be finalized based on the Reputation-based Majority Voting mechanism.
Public Review: After completing the pre-order verification process, the contributed data will enter circulation. In the screening module, users can query the metadata and provide feedback. If we find that there is too much negative feedback, we will reject this submission.
The submission process ends after the above-mentioned stages of verification are passed. If the data is consumed by more users in the future, contributors will continue to receive corresponding incentives.
Last updated
