Pith. sign in

Sparse but wrong: Incorrect l0 leads to incorrect features in sparse autoencoders.arXiv preprint arXiv:2508.16560,

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

Perplexity Can Miss SAE Feature Damage Under Quantization

cs.LG · 2026-06-02 · unverdicted · novelty 6.0

Quantization of LLMs can degrade many SAE features even when perplexity improves or stays similar, as shown by correlation measurements on frozen SAEs for Pythia-70M and Gemma-2-2B models across INT8 to INT4.

citing papers explorer

Showing 1 of 1 citing paper.

  • Perplexity Can Miss SAE Feature Damage Under Quantization cs.LG · 2026-06-02 · unverdicted · none · ref 4 · internal anchor

    Quantization of LLMs can degrade many SAE features even when perplexity improves or stays similar, as shown by correlation measurements on frozen SAEs for Pythia-70M and Gemma-2-2B models across INT8 to INT4.