Table 3 Ablation result. We compared the performance of FDE-LLM with the performance after removing CA. The best data is highlighted in Bold. FDE-LLM outperforms the performance after removing CA in every dataset.
From: Social opinions prediction utilizes fusing dynamics equation with LLM-based agents
Method | Pangmao | Jiangping | Qingdao | Dianduji | ||||
|---|---|---|---|---|---|---|---|---|
DTW\(\downarrow\) | Corr.\(\uparrow\) | DTW\(\downarrow\) | Corr.\(\uparrow\) | DTW\(\downarrow\) | Corr.\(\uparrow\) | DTW\(\downarrow\) | Corr.\(\uparrow\) | |
FDE-LLM | 0.3622 | 0.9653 | 0.3664 | 0.8950 | 0.3352 | 0.9605 | 0.3404 | 0.8842 |
FDE-LLM (Without CA) | 0.5133 | 0.9489 | 0.7530 | 0.8671 | 0.7389 | 0.8952 | 0.5708 | 0.7367 |
