Semantic Convergence: Investigating Shared Representations Across Scaled LLMs
This work addresses the problem of understanding shared representations in scaled LLMs for researchers in interpretability, though it is incremental as it builds on existing methods.
The study investigated whether language models of different scales converge on similar internal concepts, finding that middle layers of Gemma-2 models showed strong overlap in feature spaces, reinforcing universality for cross-model interpretability.
We investigate feature universality in Gemma-2 language models (Gemma-2-2B and Gemma-2-9B), asking whether models with a four-fold difference in scale still converge on comparable internal concepts. Using the Sparse Autoencoder (SAE) dictionary-learning pipeline, we utilize SAEs on each model's residual-stream activations, align the resulting monosemantic features via activation correlation, and compare the matched feature spaces with SVCCA and RSA. Middle layers yield the strongest overlap, while early and late layers show far less similarity. Preliminary experiments extend the analysis from single tokens to multi-token subspaces, showing that semantically similar subspaces interact similarly with language models. These results strengthen the case that large language models carve the world into broadly similar, interpretable features despite size differences, reinforcing universality as a foundation for cross-model interpretability.