Validating Miner Work
Masa's SN42 uses a sophisticated validation system to evaluate the work of miners, and validate that they are acting in a benevolent fashion
Explanation
Validators on the Masa Subnet utilize the Masa SDK to validate the work of miners. They randomly spot check miner data to ensure that it is accurate, and relevant to the validator's initial request. In the case of Twitter data, this means:
Randomly selecting a tweet from a miner's response
Validating that the tweet does indeed exist on Twitter
Ensuring the query requested is present in the tweet
Confirming the tweet was posted within the requested time range (today)
If a miner does not pass the spot check, they are assumed malicious and are given a score of 0.
If a miner does pass the spot check, the validator then goes through each "Tweet" of the miner's repsponse, and validates that it looks like a Tweet. Validators use an all-MiniLM-L6-v2
LLM model to confirm that each Tweet passes a similiarity test, i.e. looks similar enough to an example Tweet. Only those Tweets that pass are considered valid.
Conclusion
This validation mechanism ensures that miners are acting in a benevolent manner. It encourages miners to scrape live, accurate data and submit it to the network.
By using this approach, the Masa Network aims to create a stable, reliable ecosystem where genuine value creation is recognized and rewarded.
For more information on miner operations, refer to: