Clean First, Align Later: Benchmarking Preference Data Cleaning for Reliable LLM Alignment

Published in *Advances in Neural Information Processing Systems* (**NeurIPS**) Datasets and Benchmarks Track, 2025