Realistic hiring context reintroduces race and gender bias in LLMs despite anti-bias prompts, and ablating internal demographic directions reduces that bias.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Robustly Improving LLM Fairness in Realistic Settings via Interpretability
Realistic hiring context reintroduces race and gender bias in LLMs despite anti-bias prompts, and ablating internal demographic directions reduces that bias.