MetaCheckGPT -- A Multi-task Hallucination Detector Using LLM Uncertainty and Meta-models (2404.06948v2)
Abstract: Hallucinations in LLMs have recently become a significant problem. A recent effort in this direction is a shared task at Semeval 2024 Task 6, SHROOM, a Shared-task on Hallucinations and Related Observable Overgeneration Mistakes. This paper describes our winning solution ranked 1st and 2nd in the 2 sub-tasks of model agnostic and model aware tracks respectively. We propose a meta-regressor framework of LLMs for model evaluation and integration that achieves the highest scores on the leaderboard. We also experiment with various transformer-based models and black box methods like ChatGPT, Vectara, and others. In addition, we perform an error analysis comparing GPT4 against our best model which shows the limitations of the former.
- Gpt-4 technical report. Technical report, OpenAI.
- Amos Azaria and Tom Mitchell. 2023. The internal state of an llm knows when its lying. arXiv preprint arXiv:2304.13734.
- Generationary or “how we went beyond word sense inventories and learned to gloss”. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 7207–7221, Online. Association for Computational Linguistics.
- Detecting and mitigating hallucinations in machine translation: Model internal workings alone do well, sentence similarity even better.
- Hallucinations in large multilingual translation models.
- Looking for a needle in a haystack: A comprehensive study of hallucinations in neural machine translation.
- Deberta: Decoding-enhanced bert with disentangled attention.
- The factual inconsistency problem in abstractive text summarization: A survey.
- Simon Hughes. 2023. Cut the bull… detecting hallucinations in large language models.
- Mixtral of experts.
- Chain of natural language inference for reducing large language model ungrounded hallucinations. arXiv, cs.CL(arXiv:2310.03951).
- Truthfulqa: Measuring how models mimic human falsehoods.
- Roberta: A robustly optimized bert pretraining approach.
- Selfcheckgpt: Zero-resource black-box hallucination detection for generative large language models.
- On faithfulness and factuality in abstractive summarization. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 1906–1919, Online. Association for Computational Linguistics.
- Semeval-2024 shared task 6: Shroom, a shared-task on hallucinations and related observable overgeneration mistakes. In Proceedings of the 18th International Workshop on Semantic Evaluation (SemEval-2024), pages 1980–1994, Mexico City, Mexico. Association for Computational Linguistics.
- Distilbert, a distilled version of bert: smaller, faster, cheaper and lighter.
- Retrieval augmentation reduces hallucination in conversation. In Findings of the Association for Computational Linguistics: EMNLP 2021, pages 3784–3803, Punta Cana, Dominican Republic. Association for Computational Linguistics.
- Retrieval augmentation reduces hallucination in conversation. arXiv, cs.CL(arXiv:2104.07567).
- Just ask for calibration: Strategies for eliciting calibrated confidence scores from language models fine-tuned with human feedback. https://doi.org/10.48550/arXiv.2305.14975. ArXiv:2305.14975v2 [cs.CL].
- Llama 2: Open foundation and fine-tuned chat models.
- Mutual information alleviates hallucinations in abstractive summarization.
- A stitch in time saves nine: Detecting and mitigating hallucinations of llms by validating low-confidence generation.
- Yijun Xiao and William Yang Wang. 2021. On hallucination and predictive uncertainty in conditional language generation.
Sponsor
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
Top Community Prompts
Collections
Sign up for free to add this paper to one or more collections.