Contributors and verifiers are paid $k\%$ of the total Reward Pool at the end of every epoch. The reward per participant is dependent on their reputation at the time of submission/verification.

Dataset creator specifies the following values:

$\lambda_c \in (0, 1]$: Reward multiple for contributors

$\lambda_v \in (0, 1]$: Reward multiple for verifiers

Contributor Rewards

Let $m_i$be the number of submissions by contributor $i$ at the end of current epoch

Now for any submission $j$,

$r_{ij}$: reputation of contributor $i$ while making submission j

$C_i$ : reward for contributor $i$

$\delta_j= \begin{cases}
1 &\text{if j is accepted} \\
0 &\text{otherwise } \\
\end{cases}$

$C_i =\lambda_c\sum_{j=1}^{m_i} \delta_j r_{ij}$

Verifier Rewards

Reward Tokens

Suggested Hyperparameters

Similarly for a verifier $i$ we can define rewards $V_i$as:

$V_i = \lambda_v\sum_{j=1}^{m_i} r_{ij}$

$\delta_j$ensures that contributors are only paid out for the accepted submissions, while verifiers are paid for all of their respective submissions

Suppose there are $M$contributors and $N$verifiers that made submissions in the latest epoch. Then total rewards $R$,

$R = \sum_{i=1}^{M}C_i + \sum_{i=1}^{N}V_i$

Let $F$: total $Frac$ tokens being rewarded in the latest epoch

$\bar C_i$: $Frac$ tokens being rewarded to contributor $i$

$\bar V_i$: $Frac$ tokens being rewarded to verifier $i$

$\bar C_i = \frac{FC_i }{R}$

$\bar V_i = \frac{FV_i }{R}$

Symbol | Typical Value(s) |
---|---|

$k$

$5\%$ to $20\%$

$\lambda_c$

$1$

$\lambda_v$

$0.1\lambda_c$ to $\lambda_c$