GAMA-Bench evaluates gender-asymmetric moral framing in 1,298 conflict scenarios
November 22, 2026
GAMA-Bench measures if LLMs apply inconsistent moral standards to the same misconduct based on the actor's gender. The benchmark uses 1,298 paired scenarios involving intimate and public social conflicts to detect asymmetric response framing.
HOW THIS AFFECTS YOU
●
researcherYou can use these 1,298 paired prompts to test if your models treat male and female actors differently in conflict settings.
●
policyThis shows how models might have hidden biases in moral judgment that go beyond simple stereotypes.