Improving Open-Ended Text Generation via Adaptive Decoding
Zhu, Wenhong, Hao, Hongkun, He, Zhiwei, Ai, Yiming, Wang, Rui
–arXiv.org Artificial Intelligence
Current language models decode text token by token according to probabilistic distribution, and determining the appropriate candidates for the next token is crucial to ensure generation quality. This study introduces adaptive decoding, a mechanism that dynamically empowers language models to ascertain a sensible candidate set during generation. Specifically, we introduce an entropy-based metric called confidence and conceptualize determining the optimal candidate set as a confidence-increasing process. The rationality of including a token in the candidate set is assessed by leveraging the increment of confidence. Experimental results reveal that our method balances diversity and coherence well. The human evaluation shows that our method can generate human-preferred text. Additionally, our method can potentially improve the reasoning ability of language models.
arXiv.org Artificial Intelligence
Jun-2-2024
- Country:
- South America (0.04)
- Oceania > Australia (0.04)
- Indian Ocean (0.04)
- North America
- Dominican Republic (0.04)
- United States
- New York (0.04)
- Hawaii > Honolulu County
- Honolulu (0.04)
- California > San Diego County
- San Diego (0.04)
- Canada > Ontario
- Toronto (0.04)
- Europe
- Asia
- Africa
- Democratic Republic of the Congo (0.14)
- Gabon (0.04)
- South Africa (0.04)
- Ethiopia > Addis Ababa
- Addis Ababa (0.04)
- Genre:
- Research Report > New Finding (0.93)
- Industry:
- Government
- Military (0.92)
- Voting & Elections (0.67)
- Government
- Technology: