Train2Earn Platform Updates

From the Closed Beta of Train2Earn, our development team has been reiterating the platform pushing update after update, and implementing features that improve the whole ecosystem. These product updates have been happening very quietly in the background. Moving forward we will try our best to communicate these improvements and updates so that we can all appreciate and stay updated on the work we’re all doing.

Similarity checker for Builder submissions

Currently, the Train2Earn platform has a canonical duplicate checker that prevents builders from submitting utterances with an exact canonical form with a previous submission. Mind AI has another tool that measures the similarity among utterances. Mind AI can input a certain level of similarity and identify utterances that go beyond this set threshold. These are utterances that won’t help train the Mind AI engine.

We will integrate this tool into the platform. Whenever a builder enters a submission, the platform will check not only for a duplicate canonical form but will also check if it meets the similarity checker criteria from Mind AI. This feature will help:

1. Builders avoid submitting bad utterances that affect their earnings,
2. Validators save time from going through similar submissions and assessing if they should reject under pattern diversity or not, and
3. Architects will get better quality data faster and reduce campaign cost

This feature will be complemented with more guidelines from Mind AI to support the community.

Implementing Builder Levels

You’ll notice that there’s a “LV 1 Builder” on the top left of your Builder homepage. The main idea is that Builders will progress through levels as they improve and will gain access to more campaigns and higher rewards. The better Builder you are, the higher level and rewards you’ll have. We’ll start with a basic structure. For example in a 3 level structure:

• Level 1: new users and those with low performance rates
• Level 2: users that meet a minimum number of approved submissions and maintain a minimum approval rate
• Level 3: users that have reached and maintained a higher approval rate
• Level 4 and higher: users that are consistent with the highest approval rate

Your performance determines your level and will determine your rewards. For example, a campaign’s base reward can be at USD 0.20 per approved and the user rewards will be as follows:

• Level 1 users can get up to USD 0.08 per approved submission;
• Level 2 can get up to USD 0.12;
• Level 3 can get up to USD 0.16;
• Level 4 can get up to USD 0.20
• Furthermore, builders with approval rate of less than 20% will get USD 0 rewards
• Tiered rewards based on approval rate will still be in effect. If you were promoted to level 3 but suddenly have a below 20% approval rate, then you will get the lowest tier reward at USD 0.00 based on this example

Tote that these figures are just sample computations. There can be more or less levels and rewards might change but the general logic remains: better builders will be rewarded more

Level 1 user rewards are lowered as they are new users who are still learning the standards and are expected to make mistakes at the beginning. Those mistakes currently represent a significant portion of the cost for a client. Capping the rewards for new users and low performers offsets that cost. Users then need to work their way up and prove themselves as good builders to unlock higher rewards.

In summary, the idea of having builder levels aims to:

• provide the opportunity for new users to improve without costing the client too much
• reward and retain good builders
• develop a sustainable Train2Earn ecosystem

Aside from the Builder Levels, we will work with the guilds to ensure that their best builders will have access to more campaigns so that we can reward and retain the most skilled users. The system is being designed to reward those who focus on quality submissions and contribute the best data.

— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —

Platform Updates

The similarity checker and levels for builders are coming soon.

Below are other updates and features that have already been built, and will be implemented in the next platform update.

Korean version of Synesis Academy

A version of our Synesis Academy for our Korean users

Refresh button on the main campaign list page

Data will be cached for quicker loading when user returns to this page

Utterance sorting for Validators

Validators will be able to sort utterances by time of submission and alphabetically to further help in detecting similar submissions

Faster access to approved data for Architects

Architects will be able to download the approved crowdsourced data as soon as the campaigns are marked as “Finished” instead of waiting for “Closed” which only happens after the users claim their rewards. Basically the client can get their requested data faster!

Campaign summary for Architects

Architects will get a summary of the campaign performance and cost

The community has been essential during the development phase providing their valuable feedback. Thank you everyone for your support!

Let’s keep on building!