RESI, Bittensor’s Subnet 46, has rolled out a new incentive mechanism built around a daily model competition focused on remote real estate appraisals in the United States.
The goal is simple: produce accurate property valuations using models that can be tested, compared, and rewarded in a transparent way.
How the Competition Works
Miners publish valuation models to Hugging Face. These models predict property prices using standard listing details such as address, number of bedrooms and bathrooms, and square footage.
Validators then run these models against a dataset of recently sold properties. Each model’s predictions are compared to the actual sale prices, and performance is measured by mean error across roughly 500 properties.
The model with the lowest error takes the entire daily reward.
Only one competition runs per day. This design choice is intentional—it reduces the risk of miners hard-coding known sale prices and keeps the focus on generalizable valuation models rather than memorization.
Transparency and Tracking
Miner performance is publicly visible through the RESI dashboard, giving participants a clear view of how models are ranking and improving over time: https://dashboard.resilabs.ai/
More Than a Single Use Case
This incentive mechanism is the first in a planned series of sub-subnets. The roadmap starts with residential appraisals, then expands into:
- Commercial property appraisals
- Residential and commercial inspections
Together, these pieces are meant to support a broader framework for end-to-end real estate intelligence. The long-term vision combines model competitions, scoring systems, and additional primitives into an API that can support appraisals, inspections, title work, and legal processes.
What’s Coming Next
Upcoming updates to the subnet include:
- Adding property images to the validation dataset
- Requiring licenses for submitted models
- Introducing combined model + agent competitions
RESI is positioning itself as a practical, data-driven layer for real estate intelligence, using competitive incentives to push accuracy rather than speculation.
Be the first to comment