Key Terminology
Topics
Schelling Points (opens in a new tab) that focus the efforts of the protocol by categorizing inferences. Anyone (identified as topic creators) can permissionlessly create a topic on Allora and define a rule set that determines how to reward correct inferences within said topic. Workers can then submit inferences to these topics and earn rewards based on how accurate their inferences are.
Rule Set
Loss calculation logic determined at topic creation by a topic creator who decides how to evaluate inferences, and consists of:
- The loss function to use (e.g., mean absolute directional loss, L1-norm)
- The source of ground truth (e.g., some endpoint, some oracle, the median of gathered inferences)
Inferences
Predictions or conclusions made by workers about specific outcomes within a given topic.
Forecasts
Predictions made by workers about the performance of their peers in the current epoch, expressed as a set of forecasted losses in accordance with the topic's loss function. Forecasts are used to gauge the reliability and accuracy of the participants' inferences within the specific context provided by the current circumstances.
Forecasts and predictions are used interchangeably throughout the docs when referring to the output of forecasters.
Context Awareness
An additional dimension of evaluation that enables the network to achieve the best inferences under any circumstances using inferences and forecasts provided by workers.
By incorporating feedback from the test set (live, revealed ground truth plus the live, revealed performances of one's peers), both inference and forecast models of individual workers can be improved over time. This improves overall network performance.
By incentivizing forecasts, workers are incentivized to understand the contexts in which they and their peers perform well or poorly. For example, forecasters may understand that a subset of workers perform better on Wednesdays, whereas another performs well on Thursdays, or some do well in bear markets and others in bull markets. Integrating such context-aware insights empowers Allora to admit better performance than any individual actor because it can selectively leverage insights from the appropriate actors in the appropriate contexts.
Network Participants
A network participant in the Allora Network is an individual or entity that contributes and continuously adds value to the network by fulfilling specific roles.
Supply Side
All network participants that are not consumers. This includes workers, reputers, and validators. By contrast, the demand side is entirely informed by consumers or those who request inferences.
Epochs
Discrete periods during which inferences and forecasts are submitted, and rewards are distributed. Each epoch provides a timeframe for evaluating and scoring the performance of workers and reputers.
Rewards
Incentives given to workers and reputers based on their accuracy, performance and/or stake. These rewards are distributed at the end of each epoch, encouraging high-quality contributions.
Stake
A financial commitment made by reputers to show confidence in their ability to assess reputation by sourcing the truth and comparing it to workers' inferences. This stake increases the importance and rewards of a topic. Participants use the Allora chain CLI to stake.
Delegated Stake
A method for passive earnings where funds are delegated to a reputer, allowing the delegator to receive rewards based on the reputer's performance.
- These delegated funds enhance the reputer's stake, improving topic security and the accuracy of loss reports
- A withdrawal delay prevents quick attacks
- Delegating involves risk but offers rewards based on the reputer's success
Withdrawal Delay
Allora enforces a mandatory waiting period for withdrawals of stake and rewards to enhance security. When you request a withdrawal, you must:
- Initiate the withdrawal
- Wait for the specified delay before the withdrawal executes
This two-step process helps protect against flash attacks while maintaining a smooth user experience.
Regrets
A measure of how the performance of a worker’s inference compares to the network’s previously reported accuracy.
A positive regret implies that the inference of worker x outperforms the network, whereas a negative regret implies the network outperforms worker x.