Yuxin Jiang

Ms. Yuxin Jiang


Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

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

Representative Publications

View All Publications
19 December 2025 in Publications
Verified
Journal Article
Open access
PDF file
Zhihe YangZhihe Yang
Zhihe Yang
Xiaoyu ZhouXiaoyu Zhou
Xiaoyu Zhou
YJYuxin Jiang
Yuxin Jiang
XZXinjie Zhang
Xinjie Zhang
QGQihui Gao
Qihui Gao
AYAnqi Yang
Anqi Yang
In response to the persistent imbalance between the rapid expansion of scholarly publishing and the constrained availability of qualified reviewers, an empirical investigation was conducted to examine the feasibility and boundary conditions of employing Large Language Models (LLMs) in journal peer review. A parallel corpus of 493 pairs of human expert reviews and GPT-4o-generated reviews was constructed from the open peer-review platform PeerJ Computer Science. Analytical techniques, including keyword co-occurrence analysis, sentiment and subjectivity assessment, syntactic complexity measurement, and n-gram distributional entropy analysis, were applied to compare cognitive patterns, evaluative tendencies, and thematic coverage between human and AI reviewers. The results indicate that human and AI reviews exhibit complementary functional orientations. Human reviewers were observed to provide integrative and socially contextualized evaluations, while AI reviews emphasized structural verification and internal consistency, especially regarding the correspondence between abstracts and main texts. Contrary to the assumption of excessive leniency, GPT-4o-generated reviews demonstrated higher critical density and functional rigor, maintaining substantial topical alignment with human feedback. Based on these findings, a collaborative human–AI review framework is proposed, in which AI systems are positioned as analytical assistants that conduct structured verification prior to expert evaluation. Such integration is expected to enhance the efficiency, consistency, and transparency of the peer-review process and to promote the sustainable development of scholarly communication.

25 August 2026 in Publications
Verified
Journal Article
Open access
PDF file
Zhihe YangZhihe Yang
Zhihe Yang
Xiaoyu ZhouXiaoyu Zhou
Xiaoyu Zhou
WWWang Hongsa Wang
Wang Hongsa Wang
YJYuxin Jiang
Yuxin Jiang
XZXinjie Zhang
Xinjie Zhang
Large language models (LLMs) are increasingly employed in scholarly peer review, yet their suitability as autonomous evaluators remains uncertain. Using the ICLR 2025 review process, this study compares 2401 human reviews with 7203 reviews produced in separate, context-isolated API runs using Claude Sonnet 4.5, GPT-5.2 Thinking, and Gemini 3 Pro Preview across decision agreement, review-text characteristics, inter-model consistency, and human–AI aggregation. Raw LLM scores showed systematic leniency and score compression. A 0.1-point grid search identified thresholds of 6.2, 6.3, and 6.7 for Claude, GPT, and Gemini, respectively; repeated stratified cross-validation reproduced these thresholds. When applied without retuning to a stratified balanced sample of 300 ICLR 2024 papers, decision-agreement accuracy was 0.927, 0.913, and 0.930. Independent human coding of research type and primary field showed substantial pre-adjudication agreement (Cohen’s kappa = 0.774 and 0.714), and the recalculated analyses did not support H3. Review-text indicators showed similar structural completeness across sources but uneven critical-section length; these descriptive measures do not establish review quality. Human-containing aggregation rules showed higher agreement with conference decisions than corresponding AI-only rules, without establishing independent review quality or causal complementarity. A textual-overlap check found very low exact eight-gram containment, and manual inspection of the highest-similarity 1% found shared manuscript content or domain terminology rather than reviewer-specific evaluative language; possible prior exposure nevertheless could not be excluded.