MANTIS is a Bittensor subnet that rewards miners for embeddings which help predict financial instrument targets with a 1-hour horizon. The system uses time-lock encryption to prevent miners from seeing future prices before submitting, ensuring fair competition. Validators collect encrypted embeddings, align them with price moves, and assign weights based on salience using XGBoost aggregation and permutation importance.
This subnet is currently safe from deregistration.
No contribution data available
Yuma Pulseâ„¢
Financial Predictions
Incentivizes embeddings that predict financial instrument movements with 1-hour horizons
Time-lock Encryption
Drand beacon integration prevents miners from seeing future prices before submission
Salience Scoring
XGBoost aggregation with permutation importance determines embedding quality and rewards
Multi-challenge System
LBFGS challenges with classifier and Q-salience combined for comprehensive evaluation