A newly posted arXiv paper argues that current AI systems risk outpacing human understanding and control, and that fresh mathematical work is needed to make them legible, steerable, and cooperative. The author structures the call by mathematical subfield so researchers can identify where their expertise applies, and frames it as an open invitation to the mathematics community.
A new arXiv paper argues that benchmarks testing cooperation between AI agents may overlook the unwritten conventions that let humans infer meaning beyond literal messages. The authors introduce the concept of a "convention gap" and outline an approach for quantifying implicit communication in cooperative AI evaluations. The work is a research proposal rather than a released model or tool.
A revised arXiv preprint introduces an approach to generating cooperative multi-agent tasks and certifying when cooperation is actually required. It works within the Laser Learning Environment, a multi-agent path-finding setting where shared rewards alone do not guarantee that agents must collaborate. The method aims to formalize when, how, and whether agents need to cooperate to succeed.