GGSS: Geodesic-Gated Spherical Steering for Inference-Time Debiasing of Generative Vision-Language Models
- 类型:arxiv
- 标识:2608.25375
- 链接:https://arxiv.org/abs/2608.25375
- 主分类:llm-infra
- 形态:method
- TLDR:Generative vision-language models (VLMs) are increasingly used in human-centered settings, yet they can produce demographically biased outputs even when images differ only in controlled attributes such as perceived race or gender. However, existing inference-time debiasers were largely designed for static embeddings or CLIP-like models rather than generative VLMs. We propose GGSS---Geodesic-Gated Spherical Steering---a norm-preserving intervention that discovers a counterfactual bias subspace on the unit hypersphere, steers visual tokens along geodesic arcs, and uses an adaptive gate to focus
- 副分类:multimodal
- 待LLM分类:否
- 来源文件:
- /inbox/tom/_candidates/2026-08-31-agent-rag-longcontext-candidates.json