Abstract
We investigate the problem of generating instructions to guide humans to navigate in simulated residential environments. A major issue with current models is hallucination: they generate references to actions or objects that are inconsistent with what a human follower would perform or encounter along the described path. We develop a model that detects these hallucinated references by adopting a model pre-trained on a large corpus of image-text pairs, and fine-tuning it with a contrastive loss that separates correct instructions from instructions containing synthesized hallucinations. Our final model outperforms several baselines, including using word probability estimated by the instruction-generation model, and supervised models based on LSTM and Transformer.- Anthology ID:
- 2023.findings-emnlp.266
- Volume:
- Findings of the Association for Computational Linguistics: EMNLP 2023
- Month:
- December
- Year:
- 2023
- Address:
- Singapore
- Editors:
- Houda Bouamor, Juan Pino, Kalika Bali
- Venue:
- Findings
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 4044–4053
- Language:
- URL:
- https://aclanthology.org/2023.findings-emnlp.266
- DOI:
- 10.18653/v1/2023.findings-emnlp.266
- Cite (ACL):
- Lingjun Zhao, Khanh Nguyen, and Hal Daumé III. 2023. Hallucination Detection for Grounded Instruction Generation. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 4044–4053, Singapore. Association for Computational Linguistics.
- Cite (Informal):
- Hallucination Detection for Grounded Instruction Generation (Zhao et al., Findings 2023)
- PDF:
- https://preview.aclanthology.org/emnlp-22-attachments/2023.findings-emnlp.266.pdf