Bruno
@brunomazorra
48 Following
20 Followers
Casts
Casts + replies
Bruno
@brunomazorra
Full paper https://arxiv.org/pdf/2312.17058.pdf
0 reply
0 recast
0 reaction
Bruno
@brunomazorra
TLDR: If you impose the mechanism to be resistant to Sybil strategies, then forget about efficiency :(. If you allow agents to use Sybil strategies, the Shapley value cost-sharing mechanism is close to optimal, even when the agents have arbitrary subjective beliefs over other players actions :)!
1 reply
0 recast
0 reaction
Bruno
@brunomazorra
New paper on funding public excludable goods (or club goods) in permissionless settings.
1 reply
0 recast
2 reactions
Bruno
@brunomazorra
I tried to study concave pro-rata games, and I conjectured that the welfare was around O(2^N/N) (or something like that I do not remember XD), but I wasnt able to prove it! I will read the paper thanks!:)
0 reply
0 recast
1 reaction
Bruno
@brunomazorra
An interesting research direction would be to analyze the equilibria that emerge when agents make decisions sequentially (Stackelberg games).
1 reply
0 recast
0 reaction
Bruno
@brunomazorra
When there is no cost incurred by bidders that just win one of the blocks, separate auctions are close to efficient (see https://arxiv.org/pdf/1408.1211.pdf). I guess when there are costs, this is no longer true...
0 reply
0 recast
0 reaction
Bruno
@brunomazorra
Note: I think that 2-approximation can be improved by changing the commitments and running a VCG auction. But this is not the problem. When validators can run side-channel auctions, the problem is how to compare the revenue of running two separate auctions to the VCG one.
1 reply
0 recast
0 reaction
Bruno
@brunomazorra
More thoughts on this: Making a truthful, credible and 2-welfare approximated cross-domain auction is reasonably easy by using DRA (https://arxiv.org/abs/2301.12532) on the bundle, when sellers (validators) do not have access to side-channel/separated auctions.
1 reply
0 recast
0 reaction
Bruno
@brunomazorra
Cross-domain auctions are so **** hard :(.
1 reply
0 recast
0 reaction
