Teaching with Lies: Curriculum DPO on Synthetic Negatives for Hallucination Detection

ACL ARR 2025 May Submission4497 Authors

Abstract: Aligning large language models (LLMs) to accurately detect hallucinations remains a significant challenge due to the sophisticated nature of hallucinated content which makes it inherently difficult to identify. Recognizing that hallucinated samples typically exhibit higher deceptive quality than traditional negative samples, we use these carefully engineered hallucinations as negative examples in the DPO alignment procedure. Our method incorporates a curriculum learning strategy, gradually transitioning the training from easier samples, identified based on the greatest reduction in probability scores from independent fact checking models, to progressively harder ones. This structured difficulty scaling ensures stable and incremental learning. Experimental evaluation demonstrates that our HaluCheck models, trained with curriculum DPO approach and high quality negative samples, significantly improves model performance across various metrics, achieving improvements of upto 24\% on difficult benchmarks like MedHallu and HaluEval. Additionally, HaluCheck models demonstrate robustness in zero-shot settings, significantly outperforming larger state-of-the-art models across various benchmarks.

Paper Type: Short

Research Area: NLP Applications

Research Area Keywords: misinformation detection, hallucination detection, LLM alignment

Contribution Types: NLP engineering experiment, Publicly available software and/or pre-trained models

Languages Studied: English, Python

Submission Number: 4497