Date Presented:

24 July

摘要:

Chinese-domestic large language models such as Doubao, DeepSeek, and Qwen have become the primary LLM platforms, and have the potential to shape the digital environment for classical scholarship within Chinese digital domains due to their ubiquitousness, low-cost and ease of use. Yet their training data for ancient Greek and Latin philology is sparse relative to frontier Western models that remain regionally inaccessible. This paper reports a format-comparison experiment testing whether, and by what mechanism, the encoding format of a digital edition determines the scholarly quality of AI output when these models are queried about a disputed passage in the so-called Periplus Maris Erythraei §56 (line 537), where the phrase μεστὰ πλοῖα has attracted four competing readings. The paper tests whether, and by what mechanism, the encoding format of a digital edition can lead to good philological reasoning of complex textual phenomena. The study may bear implications for further efficient and AI-primed enrichment of digital editions within the Chinese domain that facilitates research and study of Classical Studies in the European tradition.

访问链接