Tensorplex Docs
  • Introduction
  • Tensorplex LST
    • Introduction
    • stTAO
    • Contract and Wallet Addresses
    • Security
    • Terms and Conditions
    • Fees
  • Tensorplex Stream
    • Introduction
    • Official Links
    • Disclaimer
  • Tensorplex LLM
    • Introduction
    • Guides
  • Tensorplex Dojo Subnet
    • Introduction
    • Whitepaper
    • Guide - Miner Setup
    • Guide - Validator Setup
    • Guide - Contributor Setup
    • Guide - Contributor Tasks
    • Official Links
  • Tensorplex TAO Bridge
    • How to Bridge
    • Technical Overview & Security
    • Contract Addresses
    • Fees
    • Terms and Conditions
  • stTAO Intelligence Chips
    • Guide and Rates
    • Referral Program
    • Disclaimer
Powered by GitBook
On this page
  1. Tensorplex Dojo Subnet

Guide - Contributor Tasks

PreviousGuide - Contributor SetupNextOfficial Links

Last updated 5 months ago

To help miners maximize emissions we are releasing some official guidelines on how to rate tasks on Dojo:

Miners are asked to score four responses on their ability to implement the requirements from a given specification. A good response is one that fully implements the features from the specification. Any response that perfectly and completely satisfies the specification is deserving of a maximum score.

A response that does not fully implement all features from the specifications should be penalized.

A response that implements additional features that are not specified in the question should also be penalized.

Examples

Tips to excel in Dojo subnet:

Dojo subnet is a competition. The lowest scoring miners will be de-registered when new miners request to register. When there is high volume of new registrations, there will be a corresponding number of de-registrations. To maintain registration in Dojo, miners must keep their relative rank higher than the lowest miners.

Miners that are de-registered will not be refunded their registration fee. We encourage miners to ensure that they are capable of completing the tasks when choosing to register on the subnet as the registration costs will be recycled to future emissions.

Miners can improve their rank by scoring more tasks accurately. Miners that follow the advice from this document can expect to have accurate performance. Miners that submit scores that are picked at random are unlikely to have good performance.

Another way to keep the number of task submissions high is to recruit contributors to cover all time zones. This way, your community of contributors can be always active to work on tasks that will be streamed by Validators 24/7. This can be done by distributing your subscription keys to your contributors that you have curated.

Reference Specification - Fig. 1-4 are different code responses generated from the reference specification.
Fig. 1 - This response most closely implements the specification. As such it should have the highest score among all of the responses.
Fig.2 - This response does not fully implement the specification requirements. It is lacking colour gradients and visual particles. As such it should receive a lower score when compared to Fig. 1
Fig. 3 - This response is hallucinating features that were not mentioned in the specifications and should be scored lower than Fig. 1. The temperature slider is not mentioned in the specifications.
Fig. 4 - The subject of this response (muscle contraction) does not match the subject of the specification (bone strain). It should be scored the lowest of all the responses, even though the technical requirements are well implemented.