Transcoders

Assumptions and requirements

Standard SAEs do not explain the nonlinear transformations (i.e., weight matrices of MLP blocks and attention heads) themselves that produce the activations. A transcoder achieves this by applying dictionary learning to the mapping, learning features that explain the transformation.

Intended applications

SAEs extract features from static activations and thus often suffice for residual-stream analysis. Transcoders track how features emerge from input to output and are essential for analysing MLP layers (Dunefsky et al., 2024).

On MLP outputs, transcoders are the appropriate tool: a single-layer SAE cannot span the nonlinearity, so transcoders learn the input-to-output map directly (Dunefsky et al., 2024), and Paulo et al. (2025c) reported favourable comparisons against SAEs there.

Limitations

Transcoders improve this by imitating MLP behaviour and revealing nonlinear transformations, though they require sequential analysis across depth.

Related methods

• Sparse Autoencoders (SAEs)
• Cross-Layer Transcoders (CLTs)
• Crosscoders
• Sparse Mixture of Linear Transforms (MOLT)
• Model Diffing

Compare side by side →

Run it

• Notebook: Transcoders: Bridging Nonlinear Layers · Open in Colab

Key papers

Cited in §6.6.1.

• ■ Jacob Dunefsky, Philippe Chlenski, and Neel Nanda. Transcoders find interpretable llm feature circuits. In Proceedings of the 38th International Conference on Neural Information Processing Systems, NIPS ’24, Red Hook, NY, USA, 2024. Curran Associates Inc. ISBN 9798331314385.

Other methods in Multi-Layer Feature Extraction and Analysis: Transcoders and Crosscoders

Cross-Layer Transcoders (CLTs) · Crosscoders