At the core of the entire protocol is the Credmark governance system. This governance system is controlled by CMK token holders and is a key part of enabling the game theoretic incentives to work.
Governance takes on many important roles, from preventing bad actors from abusing the model submission process to selecting key parameters and expanding the data sources list. Furthermore, they guide how the funds in the Credmark DAO are spent.
Because the role of CMK token holders is vital for establishing equilibrium, voters are rewarded with tokens for performing their duties when they vote. This incentivizes high reward turnouts, allowing the mechanisms to work naturally.
Governance responsibilities include voting for:
- Adding new data sources.
- Slashing the stake of users abusing the model submission process.
- Adding new categories.
- Grants to integrate new applications.
- Model selection strategy.
- Whitelisted accepted payment currencies.
Figure 5: Voting Example
The whitelist of data sources is key to the accuracy of results generated by the Credmark platform, as it defines the data that models consume. It’s important for this data to be as accurate and failure-proof as possible, which necessitates expanding the set of data sources. The more data sources, the higher chance that the data will be always available while also removing any outlier data that may arise due to errors in the data aggregation process by each of the sources.
To ensure that Credmark can continually access the greatest number of high quality data sources, the governance system will allow CMK holders to vote to add (or remove) data sources.
As can be seen in the example UX image above, Credmark users will have an easy-to-use web3 application interface for interacting with governance. This example specifically highlights how the UI will look for CMK holders voting to add new data sources. The relevant information for adding a new data source is included, allowing a user to expand on why adding their new data source would be a valuable addition to the platform.
Once the voting period for the proposal is finished (ex. four weeks), if more CMK holders voted for the proposal than against it (with a lower limit of required votes, ex. 20%), then it is accepted and the new data source is officially added and used.
When submitting models, contributors are required to stake their CMK tokens at the model attestation registry. The purpose of this is to prevent sybil attacks(18), which waste resources running useless models with no intention of reaching the leaderboard. Without staking, the protocol would be vulnerable to abuse.
Governance has the role of ensuring all models in the attestation registry are real, submitted by model contributors and aimed at reaching the top of the leaderboard and generating valuable results for end users.
If this isn’t the case, governance has the responsibility to vote on removing the models in question from the attestation registry. If the vote passes, the staked CMK tokens are confiscated in order to disincentivize bad actors from submitting further models that merely waste resources.
This allows for the model attestation registry to have a low barrier of entry for anyone to add their models without a centralized gatekeeper, while also keeping the protocol intact by protecting it from abuse.
The Credmark protocol has a whitelisted set of categories that are supported at a given point in time. Models submitted by model contributors must target one of the currently supported categories.
This list of categories is limited at first and expanded over time because:
- Credmark must run all models for each category, which must be scaled smoothly in tandem with demand for new categories.
- It ensures each category has all of the tooling prepared on launch (with regard to testing and devops among others) so that every model category supported by Credmark is production ready.
- With a limited number of model creators, having an infinite number of categories makes it challenging for any single category to have enough models to foster competition, which generates better accuracy.
As such, CMK holders have the opportunity to vote every month to expand the list of model categories that the web3 application supports.
Adding new categories for web3 applications that are already integrated with Credmark is generally simple, and takes little implementation work. However, this is not the case for new web3 applications.
Onboarding new web3 applications using Credmark takes initial upfront work in the form of time and developer resources. This translates to a small barrier of entry for Credmark to expand support, which is not optimal.
As such, CMK holders can also vote to provide grants from the governance DAO funds to external teams who will integrate Credmark into their web3 application. This speeds up adoption and allows for more users to get used to the benefits of Credmark. In turn, this will bring in more income from usage, and give contributors more opportunities to build new models.
Once a grant vote has passed, governance is expected to vote soon after on adding new categories for the web3 application. With the integration work running in tandem, it’s possible for model contributors to begin moving forward with model development.
Every category has a selection strategy that defines how to weight the accuracy of a model. This strategy can be improved over time, so the ability to update the model selection strategy is an important feature. CMK token holders can vote to change the model selection strategy for a given category.
Voting can aid in revising bad selection strategies, but can also be seen as a method for a given category to be more bearish or more long. In other words, it allows for a category ’s selection strategy to be pushed into a more conservative or more aggressive direction, depending on the decision of the CMK holders and whether the external crypto markets are going through a bull or a bear phase
Besides using CMK tokens to utilize the on-chain model result oracle, users will be able to use a varied set of currencies such as ETH, USDT, USDC, etc. Those alternative currencies will incur a higher cost, but will provide an easy method of integration with existing web3 applications and not create an artificial barrier of entry. Any of these tokens that are used in place of CMK will be used to buy CMK afterwards, thereby stimulating demand while simplifying UX.
The whitelist of accepted payment currencies is controlled by governance. Initially, the set of currencies will comprise the most popular currencies, but as the DeFi ecosystem grows and novel currencies come into use supporting it will be important for Credmark to support them to remain competitive.
Credmark Enhancement Proposals (CrEP) are the natural extension of the governance system. They provide a standardized and streamlined mechanism for major improvements to be proposed in a format that’s been battle tested by major blockchains like Bitcoin and Ethereum, all the way down to many popular web3 applications.
Any Credmark user can think of new ideas for improving Credmark and create a CrEP with no barriers to entry. It’s as simple as writing out the proposal at length and following the basic standard as displayed below:
Figure 6: Credmark Enhancement Proposal
The public nature of CrEPs also means that model contributors and developers have an opportunity to acquire public recognition for submitting valuable proposals that get submitted, and the chance to receive a grant to implement those proposals.
There are many potential types of CrEPs that can be created, with many ideas that are likely hard to even imagine today. The list below is a non-exhaustive set of ideas for the kinds of proposals possible:
- Updating output schemas for existing categories.
- Supporting models for new web3 applications.
- Novel normalization methods to improve the quality of the sourced data.
- Adding brand new data-sources.
- Implementing novel smart contracts that allow further decentralization of the system as a whole.
- Leaderboard improvements.
- Smart contract optimizations that lower gas fees for all users.
- New schemes that build off the latest innovations in game theory to pay out model creators more efficiently.
- New higher level protocols that amalgamate the results of multiple models to provide better assurance for end users.
- Transitioning Credmark to a rollup.
- Extending the model attestation registry to support expanded attestations with further useful data.
- Adding further transparency to the model execution platform via new tech.
These are just a handful of possible ideas in the short-to-medium term for CrEPs. DeFi users, developers, and data scientists are a creative group who will no doubt come up with even more, especially as both the Credmark and wider DeFi ecosystems develop.
CrEPs will allow Credmark to continue to grow past old design patterns, technology stacks, or even smart contracts. Thus, an open ecosystem where good ideas naturally filter through will evolve and provide longevity by continuously upgrading the protocol.