Thinking in Different Spaces: Domain-Specific Latent Geometry Survives Cross-Architecture Translation

ArXi:2603.20406v1 Announce Type: cross We investigate whether independently trained language models converge to geometrically compatible latent representations, and whether this compatibility can be exploited to correct model behavior at inference time without any weight updates. We learn a linear projection matrix that maps activation vectors from a large teacher model into the coordinate system of a smaller student model, then intervene on the student's residual stream during generation by substituting its internal state with the translated teacher representation.