Tag
This paper introduces a class of nonlinear axiomatic attribution methods for cooperative games to overcome the limitations of the linear Shapley value, which has an excessively large null space. Experimental results demonstrate the potential effectiveness of these methods in terms of inclusion AUC metric compared to Shapley value variants.
This paper proposes a behavioral measure of trust between AI agents based on costly verification in a cooperative survival game, studying trust formation, breakage, and recovery across six frontier model snapshots. It finds that models differ in trust calibration and that persistent over-verification is associated with indecision rather than safety.
CollabBench is a new benchmark for evaluating and training LLM agents in cooperative games, featuring diverse player simulation and a collaborative training paradigm. Experiments show 19.5% higher efficiency and 24.4% improved affective performance over base models.