TU Wien at SemEval-2024 Task 6: Unifying Model-Agnostic and Model-Aware Techniques for Hallucination Detection
Varvara Arzt, Mohammad Mahdi Azarbeik, Ilya Lasy, Tilman Kerl, Gábor Recski
Abstract
This paper discusses challenges in Natural Language Generation (NLG), specifically addressing neural networks producing output that is fluent but incorrect, leading to “hallucinations”. The SHROOM shared task involves Large Language Models in various tasks, and our methodology employs both model-agnostic and model-aware approaches for hallucination detection. The limited availability of labeled training data is addressed through automatic label generation strategies. Model-agnostic methods include word alignment and fine-tuning a BERT-based pretrained model, while model-aware methods leverage separate classifiers trained on LLMs’ internal data (layer activations and attention values). Ensemble methods combine outputs through various techniques such as regression metamodels, voting, and probability fusion. Our best performing systems achieved an accuracy of 80.6% on the model-aware track and 81.7% on the model-agnostic track, ranking 3rd and 8th among all systems, respectively.- Anthology ID:
- 2024.semeval-1.173
- Volume:
- Proceedings of the 18th International Workshop on Semantic Evaluation (SemEval-2024)
- Month:
- June
- Year:
- 2024
- Address:
- Mexico City, Mexico
- Editors:
- Atul Kr. Ojha, A. Seza Doğruöz, Harish Tayyar Madabushi, Giovanni Da San Martino, Sara Rosenthal, Aiala Rosá
- Venue:
- SemEval
- SIG:
- SIGLEX
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 1183–1196
- Language:
- URL:
- https://aclanthology.org/2024.semeval-1.173
- DOI:
- Cite (ACL):
- Varvara Arzt, Mohammad Mahdi Azarbeik, Ilya Lasy, Tilman Kerl, and Gábor Recski. 2024. TU Wien at SemEval-2024 Task 6: Unifying Model-Agnostic and Model-Aware Techniques for Hallucination Detection. In Proceedings of the 18th International Workshop on Semantic Evaluation (SemEval-2024), pages 1183–1196, Mexico City, Mexico. Association for Computational Linguistics.
- Cite (Informal):
- TU Wien at SemEval-2024 Task 6: Unifying Model-Agnostic and Model-Aware Techniques for Hallucination Detection (Arzt et al., SemEval 2024)
- PDF:
- https://preview.aclanthology.org/ingestion-checklist/2024.semeval-1.173.pdf