@inproceedings{thakur-etal-2025-hard,
    title = "Hard Negatives, Hard Lessons: Revisiting Training Data Quality for Robust Information Retrieval with {LLM}s",
    author = "Thakur, Nandan  and
      Zhang, Crystina  and
      Ma, Xueguang  and
      Lin, Jimmy",
    editor = "Christodoulopoulos, Christos  and
      Chakraborty, Tanmoy  and
      Rose, Carolyn  and
      Peng, Violet",
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2025",
    month = nov,
    year = "2025",
    address = "Suzhou, China",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/name-variant-enfa-fane/2025.findings-emnlp.481/",
    doi = "10.18653/v1/2025.findings-emnlp.481",
    pages = "9064--9083",
    ISBN = "979-8-89176-335-7",
    abstract = "Training robust retrieval and reranker models typically relies on large-scale retrieval datasets; for example, the BGE collection contains 1.6 million query-passage pairs sourced from various data sources.However, we find that certain datasets can negatively impact model effectiveness $\textemdash$pruning 8 out of 15 datasets from the BGE collection, reduces the training set size by 2.35$\times$, surprisingly increases nDCG@10 on BEIR by 1.0 point.This motivates a deeper examination of training data quality, with a particular focus on ``false negatives'', where relevant passages are incorrectly labeled as irrelevant.We utilize LLMs as a simple, cost-effective approach to *identify* and *relabel* false negatives in training datasets.Experimental results show that relabeling false negatives as true positives improves both E5 (base) and Qwen2.5-7B retrieval models by 0.7-1.4 points on BEIR and by 1.7-1.8 points at nDCG@10 on zero-shot AIR-Bench evaluation.Similar gains are observed for rerankers fine-tuned on the relabeled data, such as Qwen2.5-3B on BEIR.The reliability of LLMs to identify false negatives is supported by human annotation results. Our training dataset and code are publicly available."
}