Tokenomics for AI Arena: Rewarding Collaboration and Performance

AI Arena leverages the FLock framework to create a dynamic, fair, and incentivized ecosystem for participants in decentralized machine learning. By aligning rewards with performance, contribution, and community consensus, FLock ensures sustained engagement and system integrity.

The total supply of FLOCK, our native token, is capped at 1 billion tokens. Only the day-one unlocked tokens are minted initially, with additional tokens minted daily through our management smart contract. To know more about the airdrop and token allocation mechanism, visit here.

Key Features of AI Arena Tokenomics

1. Rewarding Honesty and Contribution

FLock incentivizes participants—training nodes, validators, and delegators—to contribute meaningfully. Rewards are determined by:

• Accumulative staking amounts: Stake size influences the reward pool.
• Relative performance: Ranked quality of submissions.
• Community governance: DAO voting sets key parameters like stake weight.

2. Dynamic Token Emission

Daily token emissions fuel rewards, with a flexible ratio split between AI Arena and FL Alliance. Once tasks are created, the distribution of rewards between DAO-verified AI Arena and FL Alliance tasks is dependent on their relative stake amount of active tasks. This blog post focuses only on AI Arena for now, so we will save FL Alliance for another time!

Participant Roles and Rewards

Training Nodes

Training nodes fine-tune AI tasks and ensure ecosystem health through staking and high-quality submissions.

• Reward Breakdown:

1. Reward A: Daily emitted rewards.
2. Reward B: Vested rewards released after task completion.

• Reward Drivers:

1. Relative stake in a task.
2. Delegated stake contributions.
3. Submission quality, determined by the ranking based on the consensus results of validators.

Formula: 

Step 1: Reward distribution within a single AI Arena task

Within a single AI Arena task, the reward distribution between training nodes and validators is determined based on their relative stake amounts. 

We assume there are $𝑛$ submissions $(𝑂1,...,𝑂𝑛)$ from $n$ training nodes with stakes $(𝑡1,...,𝑡𝑛)$, and $m$ validators $(𝑉1,...,𝑉𝑚)$ with stakes $(𝑠1,...,𝑠𝑚)$. Each validator $𝑉𝑗(1\le𝑗\le𝑚)$ evaluates the 𝑛 models submitted by the training nodes.

Let the total daily reward allocated to a task be denoted as R0​ and the parameter γ controls the split rewards, defining the balance between fixed and stake-dependent reward components.

The total rewards for training nodes are:

$R_0\cdot\left(\gamma+(1-2\gamma)\cdot\frac{\sum_{j=1}^ms_j}{\sum_{i=1}^nt_i+\sum_{j=1}^ms_j}\right)$

Step 2: Rewards for training nodes and their delegators

We can now compute the total rewards allocated for the training nodes as well as their delegators, which is based on the quality of their submission and their total amount of stake: 

$f_i(g_i,t_i)=\frac{g_i\cdot t_i^{\alpha_t}}{\sum_{k=1}^ng_k\cdot t_k^{\alpha_t}}$

In which 𝑡𝑖 the total stake amount from the training node 𝑖 as well as its respective delegators, gi is the scores of the submitted models from training node, whereas 𝑘 denotes a given training node’s rank amongst its peers in the same task. On theother hand, 𝛼𝑡 is a system parameter that determines the influence of the stake on the reward distribution.

Step 3: Rewards for training nodes

If a training node 𝑖’s stake in the task is 𝑡𝑛 and stakes delegated to training node 𝑖 is 𝑡𝑑 , i.e., 𝑡𝑖 = 𝑡𝑛 + 𝑡𝑑, then the actual reward for training node 𝑖 is

$f_i\cdot\left(\sigma+(1-\sigma)\cdot\frac{t_n}{t_n+t_d}\right)$

Note that in the front-end, you will see a “reward-sharing ratio”, which refers to (1 - 𝜎), which means when reward-sharing ratio is 60%, 𝜎 is 0.4. This ratio is set by training nodes and validators permissionlessly.

Example: 

Let’s assume daily total rewards for all AI Arena tasks for a given day is 309,157.68. We have 1 task with 2 nodes and 3 validators.

Nodes A and B stake 3,000 and 3,500 FLOCK respectively, while validators A, B and C stake 3,000, 6,000 and 3,000 respectively. Node A also receives an additional 1,000 FLOCK from its delegators, which brings the 𝑡𝑖 (total stake including delegated stake) to be 4,000 for Node A. For simplicity, we assume γ to be 0 in this example.

First, for this given task, total rewards for *all* training nodes are:

$R_0\times\frac{\sum_{i=1}^nt_i}{\sum_{i=1}^nt_i+\sum_{j=1}^ms_j}=309,157.68\times\frac{6500}{6500+12000}\approx108,623.7$

We can then compute the rewards for *Node A and its delegators*. We are assuming that the scores for Node A and B are 0.501435 and 0.498565 respectively. Consider αt​=1, rewards for Node A (together with delegators) are: 

$f_i(g_i,t_i)=\frac{g_i\cdot t_i}{\sum_{k=1}^ng_k\cdot t_k}=\frac{0.501435\times4000}{(0.501435\times4000)+(0.498565\times3500)}\times108,623.7=58,084$

Finally, given 𝜎=0.4, the actual rewards for *Node A alone* is:

$f_i\cdot\left(\sigma+(1-\sigma)\frac{t_n}{t_n+t_d}\right)=58,084\times\left(0.4+0.6\times\frac{3000}{4000}\right)=49,371.40$

2. Validators

Validators evaluate submissions to maintain quality and reliability. Their rewards depend on:

1. Validation quality: Alignment with majority consensus.
2. Stake size: Validator’s direct and delegated stakes.
3. Validation volume: Number of successfully validated submissions.

Formula: 

Step 1: Reward distribution within a single AI Arena task

Same as how reward distribution is calculated for training nodes as explained above, the rewards for validators in the same given AI Arena task is:

$R_0\cdot\left(\gamma+(1-2\gamma)\cdot\frac{\sum_{i=1}^nt_i}{\sum_{i=1}^nt_i+\sum_{j=1}^ms_j}\right)$

Step 2: Reward distribution for validators

If the validator finishes multiple (i.e., 𝑁 ) validation tasks, then its reward is:

$f_i\cdot F\cdot\left(\sigma+\left(1-\sigma\right)\cdot\frac{S_v}{S_v+S_d}\right)$

Specifically,

• F refers to the performance of the validation, which is calculated through off-chain consensus
• fi is the rewards for all validators
• σ is the guaranteed validator return against delegators
• Sv​ is the stake amount of this validator
• Sd is the stake amount delegated to this validator

Example: 

Continuing with our example above, total rewards for *all validators* will be:

$R_0\times\frac{\sum_{j=1}^ms_j}{\sum_{i=1}^nt_i+\sum_{j=1}^ms_j}=309,157.68\times\frac{12000}{6500+12000}\approx200,534.0$

If we assume validator A has a score of 0.472768, the the reward for * validator A only* is:

$f_i\cdot F\cdot\left(\sigma+(1-\sigma)\frac{S_v}{S_v+S_d}\right)=200,534\times0.369\times\left(0.4+0.6\times\frac{3000}{3000+0}\right)\approx73,997$

3. Delegators

Delegators support training nodes and validators by staking. Rewards are based on:

1. Stake quality: Contributions of the selected node/validator.
2. Stake size: Amount staked by the delegator.

Formula: 

If a delegator has delegated Sd​ to the above-mentioned Node A, then the daily return for the delegator in this task is:

$f_i\cdot(1-\sigma)\cdot\frac{t_d}{t_n+t_d}$

Similarly, if a delegator delegates to a validator, the reward will be:

$f_i\cdot(1-\sigma)\cdot\frac{s_d}{s_v+S_d}$

Example: 

Continuing the example for Node A, for a given delegator delegated 1,000 FLOCK to Node A, the reward for *this delegator alone* is:

$f_i\cdot(1-\sigma)\cdot\frac{t_d}{t_n+t_d}=58,084\times\left(0.6\times\frac{1000}{3000+1000}\right)\approx8,712.6$

Note that rewards in delegation pools are time-weighted to balance fairness for long-term participants and incentivize new delegations. As pools grow, rewards stabilize, promoting sustained engagement. Also, delegators must maintain their stake for at least 24 hours before un-delegating. This ensures meaningful contributions and prevents exploitative behaviors.

Parameters like reward splits (γ) are fine-tuned through DAO voting. This democratized control keeps the ecosystem adaptive and equitable.

To participate in and know more about AI Arena

To participate in AI Arena, navigate to our website. If you are interested in learning more about FLOCK’s tokenomics, visit our docs or whitepaper. 

Our smart contracts underpinning this system are open-source and auditable. Explore the details here.