Evaluating the Smooth Control of Attribute Intensity in Text Generation with LLMs
Published in Findings of the Association for Computational Linguistics (ACL), 2024, 2024
Current methods for controlling LLM text generation often lack fine-grained precision. We introduce a robust framework to evaluate how “smoothly” models can control attribute intensity (e.g., varying the level of positivity in a sentence). Our findings provide a clearer understanding of the capabilities and limitations of existing control techniques.
Recommended citation: Shang Zhou, Feng Yao, Chengyu Dong, Zihan Wang, Jingbo Shang. (2024). “Evaluating the Smooth Control of Attribute Intensity in Text Generation with LLMs.” In Findings of the Association for Computational Linguistics: ACL 2024.
Recommended citation: Shang Zhou*, Feng Yao*, Chengyu Dong, Zihan Wang, Jingbo Shang. (2024). "Evaluating the Smooth Control of Attribute Intensity in Text Generation with LLMs." In Findings of the Association for Computational Linguistics: ACL 2024.
Download Paper