Toolscaler: Scalable Generative Tool Calling via Structure-Aware Semantic Tokenization
Published in Findings of the Association for Computational Linguistics: EMNLP 2025, 2025
This paper studies efficient methods for enabling LLMs to invoke large-scale toolkits. It develops a structure-aware semantic tokenization scheme and a post-guided iterative training strategy, unifying tool representation, retrieval and calling within a single generative process.
ACL Anthology | DOI: 10.18653/v1/2025.findings-emnlp.30
Recommended citation: Yunyue Su, Zhang Jinshuai, Bowen Fang, Wen Ye, Jinghao Zhang, Bowen Song, Weiqiang Wang, Qiang Liu, Liang Wang. (2025). "Toolscaler: Scalable Generative Tool Calling via Structure-Aware Semantic Tokenization." Findings of the Association for Computational Linguistics: EMNLP 2025, pp. 556-578.
Download Paper