A scalable Aumann-Shapley attribution method for million-agent systems reveals that small-scale samples structurally misattribute emergence under nonlinear macro indicators, as shown by the Attribution Scaling Bias theorem.
Title resolution pending
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
representative citing papers
Foundation models are large adaptable AI systems with emergent capabilities that offer broad opportunities but carry risks from homogenization, opacity, and inherited defects across downstream applications.
Simulations show information overload decreases source localization effectiveness in networks, with Erdős-Rényi graphs more resilient than Barabási-Albert ones and a reversal where less dense networks perform better under strong overload.
citing papers explorer
-
Attributing Emergence in Million-Agent Systems
A scalable Aumann-Shapley attribution method for million-agent systems reveals that small-scale samples structurally misattribute emergence under nonlinear macro indicators, as shown by the Attribution Scaling Bias theorem.
-
On the Opportunities and Risks of Foundation Models
Foundation models are large adaptable AI systems with emergent capabilities that offer broad opportunities but carry risks from homogenization, opacity, and inherited defects across downstream applications.
-
Nonlinear dynamics of information overload: Impact on source localization in complex networks
Simulations show information overload decreases source localization effectiveness in networks, with Erdős-Rényi graphs more resilient than Barabási-Albert ones and a reversal where less dense networks perform better under strong overload.