Author

Hongsheng Wang

h-index 2 View on Semantic Scholar

We have 1 of 5 papers

We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.

Not the right person? Other researchers publish under this name.

#natural language process... Preprint Aug 2026

Strengthening Target-Language Features: SAE-Based Steering for Multilingual Inference

An inference-time multilingual steering method that uses pretrained sparse autoencoders to identify and strengthen target-language-related features that are decoded into steering signals and injected into the model's hidden states without additional training is proposed.

Hongsheng Wang, Philipp Koehn · 0 citations