BIT-0008 EVM Weight Verification#19
Conversation
| If `E_commit` was not a Verification Epoch, these steps are skipped. | ||
|
|
||
| > [!NOTE] | ||
| > The penalty is applied directly to validator weights by zeroing them out, rather than modifying derived values like ranks or trust. This ensures that unverified weights are completely ignored by the network, as they provide no value without verification. The validator effectively becomes a non-participant for that epoch, which is the most direct and logical consequence of failing verification. |
There was a problem hiding this comment.
How about we burn their collateral an deregister their uid?
There was a problem hiding this comment.
We shouldn't, at least initially, nuke these weights completely, but multiplyying stake by 0.9 or 0.8 would be a good start for the testing period.
There was a problem hiding this comment.
We shouldn't, at least initially, nuke these weights completely, but multiplyying stake by 0.9 or 0.8 would be a good start for the testing period.
That sounds good. What do you think about hyperparameters such as the following?
| Hyperparameter | Access | Description |
|---|---|---|
weight_verification_multiplier |
sudo, root | The stake multiplier controlling the impact a failed weight verification has on the weights of an unverified validator |
weight_verification_max_multiplier |
root | The global maximum value for weight_verification_multiplier to allow for a controlled rollout of the feature. |
This makes it somewhat configurable, so globally as the feature is rolled out, gov can ramp up the latitude subnet owners are allowed in terms of max impact. Then, per-subnet as subnet owners try the feature they can ramp up the multiplier individually until they're fully confident in the contract.
2 participants
As a component of the broader Proof of Weights initiative (#10), a mechanism is required to verify that validator weights submitted to the blockchain are valid per a subnet-owner-controlled verification contract.
This BIT proposes a mechanism to enable subnet owners to deploy custom verification contracts that validate weight submissions before they affect consensus. The proposed optimistic verification mechanism maintains network security while minimizing on-chain computational overhead.
