Perturb scales models, launches dashboard soon

Perturb is prioritizing validation of its evaluation mechanism before scaling larger models. The team is rolling out updates including a larger AI model integration, adjustments to core evaluation parameters (HISTORY_SIZE, MIN_PROCESSED_COUNT, K_MINERS), and an upcoming dashboard launch. They are also exploring a potential winner-take-all mechanism to improve miner participation incentives.

• •Larger AI model integration and evaluation parameter adjustments rolling out
• •Dashboard and Weights & Biases integration coming in next update
• •Winner-take-all mechanism under evaluation for miner incentives
• •Current 60-second timeout questioned; potential reduction to 5-10 seconds discussed