Harsher on Male? Evaluating LLMs on Gender-Asymmetric Moral Framing Across Diverse Conflict Scenarios
Summary
This arXiv CS.CL paper introduces GAMA-Bench, a gender-mirrored benchmark with 1,298 scenarios to study gender-asymmetric moral framing in LLMs. It finds male actors receive more punitive, escalatory, and blame-centered framing, while female actors receive more therapeutic, empathy-oriented responses, across model families and scales. The authors provide code for reproduction on GitHub.