GAMA-Bench Paper: LLMs Show Consistent Gender-Asymmetric Moral Framing

alex_verem · x · 2026-11-23

This is the source paper behind the viral thread on models being harsher on men: arXiv:2606.14068, "Harsher on Male? Evaluating LLMs on Gender-Asymmetric Moral Framing Across Diverse Conflict Scenarios" (under review).

GAMA-Bench contains 1,298 gender-mirrored scenarios built via controlled grids and cross-model review, with a structured protocol measuring how models allocate punishment, empathy, escalation, instruction, and blame. Experiments on 10 LLMs reveal a consistent male-disadvantaging asymmetry across model families, scenario tracks, model scale, and explicit reasoning. Official code is open-sourced.

Related event: Study Finds LLMs Judge Men More Harshly for Same Misdeeds(2 posts)→

Original post →

More from Research

Research channel →