An evaluation framework and 343-task benchmark show that four code LLMs produce socially biased code, and iterative bias feedback reduces measured bias substantially.
M.; Sarro, F.; and Harman, M
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Bias Unveiled: Investigating Social Bias in LLM-Generated Code
An evaluation framework and 343-task benchmark show that four code LLMs produce socially biased code, and iterative bias feedback reduces measured bias substantially.