Tianfang Xie

Dr. Tianfang Xie


Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
26 September 2024 in arXiv
Verified
Preprint
Open access
PDF file
TXTianfang Xie
Tianfang Xie
TLTianjing Li
Tianjing Li
WZWei Zhu
Wei Zhu
WHWei Han
Wei Han
YZYi Zhao
Yi Zhao
Due to their substantial sizes, large language models (LLMs) are typically deployed within a single-backbone multi-tenant framework. In this setup, a single instance of an LLM backbone must cater to multiple users or tasks through the application of various parameter-efficient fine-tuning (PEFT) models. Despite the availability of numerous effective PEFT techniques such as LoRA, there remains a need for a PEFT approach that achieves both high efficiency during inference and competitive performance on downstream tasks. In this research, we introduce a new and straightforward PEFT methodology named \underline{P}rompt D\underline{E}pen\underline{D}ent \underline{R}epresentation M\underline{O}dification (PEDRO). The proposed method involves integrating a lightweight vector generator into each Transformer layer, which generates vectors contingent upon the input prompts. These vectors then modify the hidden representations created by the LLM through a dot product operation, thereby influencing the semantic output and generated content of the model. Extensive experimentation across a variety of tasks indicates that: (a) PEDRO surpasses recent PEFT benchmarks when using a similar number of tunable parameters. (b) Under the single-backbone multi-tenant deployment model, PEDRO exhibits superior efficiency compared to LoRA, indicating significant industrial potential.

23 March 2026 in Preprints.org
Verified
Preprint
Open access
PDF file
TXTianfang Xie
Tianfang Xie
HGHaixin Guan
Haixin Guan
This study examines structural patterns in the governance of non-normative gendered bodies in China across distinct historical and contemporary contexts. Rather than asserting a direct causal lineage, we analyze recurring regulatory mechanisms that characterized the governance of imperial eunuchs and are discernibly present in the contemporary governance of transgender individuals, two populations separated by time and circumstance but subject to comparable logics of moral classification, familial surveillance, and institutional scripting. We introduce Confucian biopolitics as an analytical lens to describe governance that operates through li (ritual propriety) and xiao (filial piety) rather than solely through juridical or biomedical systems. This lens does not propose a new totalizing theory; rather, it reorients existing biopolitical frameworks toward the kinship-based, morally-inflected power structures that have historically characterized Chinese governance of bodies. Drawing on three complementary methods (archival analysis of Qing legal and administrative sources, exploratory GIS spatial analysis of institutional distributions, and computational text analysis of Ming Shilu chronicles and Chinese Wikipedia articles on transgender and LGBT topics), we identify structural regularities rather than claim continuity of intent or transmission. Comparative cases from the Ottoman Empire and Tokugawa Japan help distinguish China-specific features from more general patterns. Findings indicate that both eunuchs and transgender individuals were governed through mechanisms centered on familial legitimacy, bodily classification, and moral-ritual normalization. Spatial analysis of contemporary LGBTQ+ NGOs reveals significant geographic clustering (I = 0.735, p < .001) in coastal metropolitan areas, mirroring the spatial isolation mechanism identified in imperial eunuch governance, with Beijing functioning as the dominant institutional hub. Computational text analysis identifies discursive resonance: the term 妖 (demon/monster, χ2(1) = 0.07, p = .788, n.s.) and 欺骗 (deception, p = .863, n.s.) appear at statistically indistinguishable rates across six centuries of text, while terms for psychiatric pathology (精神病, p < .001) and discrimination (歧视, p < .001) are exclusively contemporary—evidence of diagnostic translation within a persistent moral register. The study contributes to gender history, comparative sociology, and digital humanities by offering a multi-method framework for identifying structural recurrences in governance logics across historical ruptures. We conclude with policy implications focused on familial and linguistic mechanisms of stigma, which historical analysis identifies as the most durable sites of governance.