Machine learningML & AI Methods
Transformer Circuit Faithfulness Metrics are not Robust
The authors explore the difficulties in evaluating the performance of neural network 'circuits', emphasizing the sensitivity of current methods to changes in the ablation methodology and the need for clearer claims about circuits.
Featured in No. 57 on 17 Jul 2024 · 6 days after release · 19 citations today
- Released
- 11 Jul 2024
- First featured
- No. 57 · 17 Jul 2024
- Citations (Semantic Scholar)
- 19
- Influential citations
- 2
- Published in
- Not yet, as far as Semantic Scholar knows
- Shares when featured
- 9
- Identifier
- arXiv:2407.08734
Citations and venue from Semantic Scholar (ODC-BY), refreshed weekly. Summary: Quant Letter (CC BY 4.0).