Research questionHow can speech enhancement adapt to mismatched deployment acoustics without labeled target audio?Speech enhancement models can lose quality when deployment noise differs from their training conditions. Test-time adaptation must therefore respond to acoustic mismatch without labeled target examples.