Salvato in:
| Autori principali: | , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2508.18763 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915463587430400 |
|---|---|
| author | Hao, Chao Wang, Zezheng Huang, Yanhua Xu, Ruiwen Niu, Wenzhe Liu, Xin Yu, Zitong |
| author_facet | Hao, Chao Wang, Zezheng Huang, Yanhua Xu, Ruiwen Niu, Wenzhe Liu, Xin Yu, Zitong |
| contents | This paper investigates the enhancement of reasoning capabilities in language models through token-level multi-model collaboration. Our approach selects the optimal tokens from the next token distributions provided by multiple models to perform autoregressive reasoning. Contrary to the assumption that more models yield better results, we introduce a distribution distance-based dynamic selection strategy (DDS) to optimize the multi-model collaboration process. To address the critical challenge of vocabulary misalignment in multi-model collaboration, we propose the concept of minimal complete semantic units (MCSU), which is simple yet enables multiple language models to achieve natural alignment within the linguistic space. Experimental results across various benchmarks demonstrate the superiority of our method. The code will be available at https://github.com/Fanye12/DDS. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_18763 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Dynamic Collaboration of Multi-Language Models based on Minimal Complete Semantic Units Hao, Chao Wang, Zezheng Huang, Yanhua Xu, Ruiwen Niu, Wenzhe Liu, Xin Yu, Zitong Artificial Intelligence This paper investigates the enhancement of reasoning capabilities in language models through token-level multi-model collaboration. Our approach selects the optimal tokens from the next token distributions provided by multiple models to perform autoregressive reasoning. Contrary to the assumption that more models yield better results, we introduce a distribution distance-based dynamic selection strategy (DDS) to optimize the multi-model collaboration process. To address the critical challenge of vocabulary misalignment in multi-model collaboration, we propose the concept of minimal complete semantic units (MCSU), which is simple yet enables multiple language models to achieve natural alignment within the linguistic space. Experimental results across various benchmarks demonstrate the superiority of our method. The code will be available at https://github.com/Fanye12/DDS. |
| title | Dynamic Collaboration of Multi-Language Models based on Minimal Complete Semantic Units |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2508.18763 |