pith. sign in

Ravel: Evaluat- ing interpretability methods on disentangling language model representations

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it

fields

cs.LG 1

years

2026 1

verdicts

UNVERDICTED 1

representative citing papers

Are Sparse Autoencoder Benchmarks Reliable?

cs.LG · 2026-05-18 · unverdicted · novelty 6.0

An audit of SAEBench reveals that Targeted Probe Perturbation and Spurious Correlation Removal metrics fail reliability tests and should not be used to evaluate sparse autoencoders.

citing papers explorer

Showing 1 of 1 citing paper.

  • Are Sparse Autoencoder Benchmarks Reliable? cs.LG · 2026-05-18 · unverdicted · none · ref 15

    An audit of SAEBench reveals that Targeted Probe Perturbation and Spurious Correlation Removal metrics fail reliability tests and should not be used to evaluate sparse autoencoders.