arXiv cs.CL· Ziliang Qiu, Renfen Hu, Yi Ding, Ted Underwood·· 11 小时前AI 评分42
用多维视角重新审视 LLM 同质化问题
Revisiting LLM Homogeneity Through a Multidimensional Lens
AI 导读
研究用覆盖语义、词汇、句法、结构、风格与呈现五个维度的 25 项可解释指标重新衡量 LLM 同质化，发现嵌入向量相似度与人类语义判断的相关性（ρ=0.417）低于句子级匹配（ρ=0.606），与形式判断则无显著相关（ρ=0.140）。
正文 · 原文
Abstract:Concerns about LLM homogeneity are typically quantified with embedding similarity, which reduces two texts to a single score. We revisit homogeneity with 25 interpretable metrics spanning five dimensions: semantic, lexical, syntactic, structural, and stylistic & presentational. A human study shows that perceived similarity depends on both content (what responses say) and form (how they say it), yet embedding similarity correlates less with human meaning judgments than sentence-level matching ($\rho=0.417$ vs. $0.606$) and not significantly with form judgments ($\rho=0.140$). Across 18 contemporary LLMs, models say similar things in different words: they share global semantics and broad linguistic profiles but remain distinct in lexical characteristics, structural patterns, and interactional conventions. Across releases from 2024 to 2026, models become more like themselves but not more like one another: among significantly changing metrics, intra-model distances decrease for 17 of 22, and inter-model distances increase for 14 of 21. Across two post-training pipelines, model-specific conventions emerge during SFT and are selectively reshaped by preference optimization. These results suggest that homogeneity is multidimensional and evolving, and that embedding similarity alone can overstate cross-model homogeneity while concealing diverging intra- and inter-model trends.
| Comments: | 26 pages, 9 figures |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.01549 [cs.CL] |
| (or arXiv:2611.01549v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.01549 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Ziliang Qiu [view email]
[v1]
Sun, 1 Nov 2026 04:20:16 UTC (736 KB)
来源：arXiv cs.CL · arxiv.org