RESEARCHMonitorNEXT 12 MONTHS
One Example Is Enough to Pass Fairness Benchmarks: Rethinking Fairness Evaluation for Aligned LLMs
arXiv cs.CL — Computation and Language
Factual evidence
What the source reports
Research demonstrates that base LLMs can pass standard fairness benchmarks like BBQ using as little as a single training example.