GGSS (Geodesic-Gated Spherical Steering) is an inference-time debiasing method for generative vision-language models that discovers a counterfactual bias subspace on the unit hypersphere, steers visual tokens along geodesic arcs, and uses an adaptive gate to focus correction on tokens carrying stronger demographic signal. Evaluated against existing debiasing baselines across multiple generative VLMs, it achieves stronger bias reduction while preserving visual-language performance.
