EAP-GP adapts the integration path in edge attribution patching to avoid gradient saturation, improving circuit faithfulness on GPT-2 models.
Identifying and adapting transformer-components responsible for gender bias in an english language model
1 Pith paper cite this work, alongside 1 external citations. Polarity classification is still indexing.
1
Pith paper citing it
1
external citations · OpenAlex
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
EAP-GP: Mitigating Saturation Effect in Gradient-based Automated Circuit Identification
EAP-GP adapts the integration path in edge attribution patching to avoid gradient saturation, improving circuit faithfulness on GPT-2 models.