Introduction

Premier Infrastructure Layer for Human Intelligence

Tensorplex Dojo Subnet is an open platform designed to crowdsource diverse, unbiased human intelligence to be accessible for AI/ML development and AI agent operations both on-chain and in the real world, while enabling contributors of all technical levels to contribute to Bittensor and Decentralised AI.

We believe Decentralised AI is more than a technological innovation — it's a movement meant to counteract the natural centralizing tendencies of AI development which may be a big threat to human liberty and diversity of thoughts. However, this movement faces two critical challenges:

1. A growing community of believers lacks meaningful ways to contribute beyond advocacy and investment.

2. Decentralized and Open-Source AI desperately needs human input to thrive and compete with centralized alternatives.

This misalignment represents not just a missed opportunity, but a potential threat to the future of AI diversity. From ensuring the integrity and human-alignment of AI outputs, to generating high-quality, verifiable data, to bridging AI with real-world tasks—the need for human contributors is both urgent and largely unaddressed.

Powered by Bittensor, the Dojo Subnet allows contributors of all technical levels to contribute to Decentralized AI and earn TAO by labelling data or providing human-preference data. However, opening up the platform for anyone to access introduces a new set of challenges, such as performing quality control, verifying that contributions are genuinely human and preventing sybil attacks.

To address these challenges, Dojo introduces multiple novel features to solve this issue:

• Modular Pipelines for Synthetic Task Generation:

  • Developer-friendly, plug-and-play modules for anyone to generate unique synthetic tasks with state-of-the-art Large Language Models (LLMs) to collect human feedback data through Dojo, which will be used to improve open-source models or train specialised agents.

• Partial Ground Truth Validation Mechanism:

  • The validators will be able to synthetically generate partial ground truths, allowing them to determine the quality of responses provided by individual contributors while taking general consensus into account.

• [Coming soon] Contributor Feedback Loop:

  • An adversarial-resistant Contributor Feedback Loop where various Contributors independently provide effective feedbacks that will recursively improve the Task responses for subsequent rounds of evaluation until the convergence onto optimal Task response is achieved. The mechanism is designed to build a constantly-improving data pipeline that rewards top contributors who provide meaningful and actionable feedbacks that is also agreed by other contributors during their independent evaluations.

There are multiple ways to participate in Dojo;

• Validator:

  • Users with large amounts of TAO delegated to their address can join as a validator (Guide - Validator) to generate synthetic tasks or distribute organic tasks to the miners.

• Miner:

  • Users who would like to participate directly can spin up a miner (Guide - Miner). Keep in mind that you will need to burn/recycle a certain amount of TAO in order to join the subnet.

• Contributor:

  • Users who do not want to spin up a miner but would like to contribute through completing the tasks will need to work with a miner in order to get their subscription key (Guide - Contributor). Please also refer to Guide - Contributor Tasks to learn how to maximise your rewards.