Abstract
To be included into chatbot systems, Large language models (LLMs) must be aligned with human conversational conventions. However, being trained mainly on web-scraped data gives existing LLMs a voice closer to informational text than actual human speech. In this paper, we examine the effect of decoding methods on the alignment between LLM-generated and human conversations, including Beam Search, Top K Sampling, and Nucleus Sampling. We present new measures of alignment in substance, style, and psychometric orientation, and experiment with two conversation datasets. Our results provide subtle insights: better alignment is attributed to fewer beams in Beam Search and lower values of P in Nucleus Sampling. We also find that task-oriented and open-ended datasets perform differently in terms of alignment, indicating the significance of taking into account the context of the interaction.- Anthology ID:
- 2024.wassa-1.22
- Volume:
- Proceedings of the 14th Workshop on Computational Approaches to Subjectivity, Sentiment, & Social Media Analysis
- Month:
- August
- Year:
- 2024
- Address:
- Bangkok, Thailand
- Editors:
- Orphée De Clercq, Valentin Barriere, Jeremy Barnes, Roman Klinger, João Sedoc, Shabnam Tafreshi
- Venues:
- WASSA | WS
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 273–279
- Language:
- URL:
- https://aclanthology.org/2024.wassa-1.22
- DOI:
- Cite (ACL):
- Shaz Furniturewala, Kokil Jaidka, and Yashvardhan Sharma. 2024. Impact of Decoding Methods on Human Alignment of Conversational LLMs. In Proceedings of the 14th Workshop on Computational Approaches to Subjectivity, Sentiment, & Social Media Analysis, pages 273–279, Bangkok, Thailand. Association for Computational Linguistics.
- Cite (Informal):
- Impact of Decoding Methods on Human Alignment of Conversational LLMs (Furniturewala et al., WASSA-WS 2024)
- PDF:
- https://preview.aclanthology.org/nschneid-patch-4/2024.wassa-1.22.pdf