Skip to content
Attune
All pages

Overview

Start here

Attune, built by Robotensor, is Bittensor subnet 58. Its emission pays for two projects: Vector, an open competition for robot models that adapt to new work, and Horizon, where people collect the data robots learn from.

2 min read

Adaptive robot intelligence

Robots meet work that keeps changing: new objects, new layouts, new products. The usual answer is to collect data and retrain. Attune pays for the alternative, adaptive robot models: models that are shown the work at deployment and do it, with their weights frozen. It pays for both halves of that: the models, and the human data they learn from.

The projects

ProjectWhat it pays forWho takes partShare of miner emission
VectorModels that learn a task from one demonstration and do it in a new scene, settled in public duelsModel builders, who submit weights30%
HorizonRobot data: robot control in simulation with a phone as the controller, and first-person recordings of everyday workAnyone with a phone and a computer70%

Each project has a site of its own, with its own docs: Vector at vector.robotensor.ai, Horizon at horizon.robotensor.ai. These pages cover what is true for the whole subnet: where its emission goes, and how to validate.

How the emission moves

Validators set the same weights from the same rules, written in the subnet's code. Vector's 30% goes to its four most recent champions. Horizon's 70% goes to one hotkey, the vault, which pays contributors every week by the points their work earned. A share with no one to take it burns. Emission and the split has the rules, and The Horizon vault follows the vault's alpha to the people it pays.

The chain's own numbers (every UID, its weight and its emission) are on Taostats. This site shows none of its own.

Three ways in

  1. Compete. Train a model for Vector and commit it on chain: Vector's docs.
  2. Contribute. Control robots and record your work with Horizon; no hotkey needed: horizon.robotensor.ai.
  3. Validate. Run the subnet's validator on a GPU host: Running a validator.

Where to go next