For instance, within the scenario illustrated within the determine above, the shared sequence of tokens is perhaps “Mechanism Design for”. The distributions is perhaps [(“Large”, 0.8), (“Generative”, 0.2)] for the LLM of Agent 1, (“Massive”, 1.0) for the LLM of Agent 2, and (“Generative”, 1.0) for the LLM of Agent 3. The bids is perhaps 1, 2, and a pair of, respectively. A attainable aggregated distribution could be the bid-weighted common of the distributions, particularly [(“Large”, 0.56), (“Generative”, 0.44)]. A attainable selection for the funds could be to ask every agent to pay their bid, which might have the brokers commit 1, 2, and a pair of, respectively.
For our theoretical evaluation of this mannequin (and choices of distribution aggregation features and fee features), we assume that the brokers honestly report their distributions, however could also be strategic about their bids. We imagine it is a life like assumption, as LLMs encode preferences over output textual content in a succinct and non-obvious method. Furthermore, to ensure that the token public sale to have the ability to combination distributions, we have to have (a minimum of) some (minimal) details about agent’s preferences away from their “most well-liked” distributions. Our strategy right here is to imagine that the brokers have (identified) partial choice orders over distributions. That’s, we assume that brokers might be able to rank some, however not all, pairs of distributions.
