Aligning Sentence Embeddings to Human Concepts via Sparse Autoencoders
For practitioners of neural information retrieval, this provides a method to improve interpretability and steerability of RAG systems, though it is an incremental application of existing SAE techniques.
This work uses Top-k Sparse Autoencoders to disentangle dense sentence embeddings into human-interpretable concepts, enabling activation steering for re-ranking search results to align with user constraints without retraining.
Dense sentence embeddings are fundamental to modern Retrieval-Augmented Generation (RAG) systems but suffer from a lack of interpretability due to feature superposition. This opacity hinders the alignment of retrieval processes with human intent, as the entangled representations are difficult to analyze or control. In this work, we propose a method to disentangle the dense representations of sentence transformers (e.g., E5) into human-interpretable concepts using Top-k Sparse Autoencoders (SAEs). We demonstrate that these disentangled features align with specific semantic, syntactic, and pragmatic categories. Furthermore, we introduce an activation steering mechanism that allows for precise intervention in the retrieval process. By clamping specific latent features, we show that it is possible to re-rank search results to better align with user constraints without retraining the backbone model. Our findings suggest that SAE-based decomposition offers a viable path toward transparent and steerable neural information retrieval.