Search Results for author:
Found 6 papers, 2 papers with code
Interpretability in Parameter Space: Minimizing Mechanistic Description Length with Attribution-based Parameter Decomposition
We introduce Attribution-based Parameter Decomposition (APD), a method that directly decomposes a neural network's parameters into components that (i) are faithful to the parameters of the original network, (ii) require a minimal number of components to process any input, and (iii) are maximally simple.
Towards evaluations-based safety cases for AI scheming
We sketch how developers of frontier AI systems could construct a structured rationale -- a 'safety case' -- that an AI system is unlikely to cause catastrophic outcomes through scheming.
Identifying Functionally Important Features with End-to-End Sparse Dictionary Learning
We propose end-to-end (e2e) sparse dictionary learning, a method for training SAEs that ensures the features learned are functionally important by minimizing the KL divergence between the output distributions of the original model and the model with SAE activations inserted.
Using Degeneracy in the Loss Landscape for Mechanistic Interpretability
We propose that if we can represent a neural network in a way that is invariant to reparameterizations that exploit the degeneracies, then this representation is likely to be more interpretable, and we provide some evidence that such a representation is likely to have sparser interactions.
The Local Interaction Basis: Identifying Computationally-Relevant and Sparsely Interacting Features in Neural Networks
We present a novel interpretability method that aims to overcome this limitation by transforming the activations of the network into a new basis - the Local Interaction Basis (LIB).
Interpreting Neural Networks through the Polytope Lens
Previous mechanistic descriptions have used individual neurons or their linear combinations to understand the representations a network has learned.
