spectralbranding/meaningfulness-cross-language-rendering
收藏资源简介:
该数据集包含多语言渲染和提取工件,用于演示Zharnikov (2026ap)论文《相同意义,不同文本:组织知识工作中的脊柱保存与渲染等价性》中的命题P4(在脊柱保存下的渲染等价性)。记录捕获了将共享的组织知识脊柱渲染成英文、俄文和中文文本的过程,以及用于计算渲染等价性分数(Rec)的脊柱重新提取。五个LLM模型涵盖三个训练语料库家族——美国专有API模型、中国专有API模型和通过Ollama(Qwen3.6:27b)本地部署的开源权重模型——以测试P4在不同语言和训练语料来源下的不变性。
This dataset contains the multi-language rendering and extraction artifacts demonstrating Proposition P4 (rendering-equivalence under spine-preservation) from Zharnikov (2026ap), *Same Meaning, Different Prose: Spine Preservation and Rendering Equivalence in Organizational Knowledge Work*. The records capture renderings of a shared organizational-knowledge spine into English, Russian, and Chinese prose, plus the spine re-extractions used to compute rendering-equivalence scores (Rec). Five LLMs span three training-corpus families — proprietary US-API models, proprietary Chinese-API models, and open-weights local deployment via Ollama (Qwen3.6:27b) — to test invariance of P4 across language and across training-corpus provenance.




