AI agent which uses Zkml to predict the performance of validators and operators and reduce slashing risks.
Staking is one of the primary methods for generating income for tokens such as ETH. Currently, about 60% of ETH is used inside of LST/LRT protocols. This means that the risk of slashing is becoming increasingly significant for Ethereum tokenomics.
However, predicting such events is still an unresolved issue that could help act proactively and alert node operators to potential critical decreases in their performance score.
We propose predictions for performance of validators and operators (Eigenlayer protocol) using zkml.
The project uses Giza tech to transpile model into cairo program deployed on starknet and dataset of beacon chain validators and eigenlayer operators from lambda (p2p.org).
We used Linear Regression and xgboost to train the model. The dataset is timeseries and we calculated past seven days effectiveness to train the model to predict effectiveness of a validator.