Primary sourcearXiv: OpenAI
View sourceEvent ·Official first-hand
KinyaEmbed:基尼亚万达语对比句嵌入模型
KinyaEmbed是首个专为基尼亚万达语设计的句子嵌入模型,该语言在卢旺达有1200万使用者。模型基于KinyaBERT-large,通过四阶段课程训练:第一阶段使用约18,000对同义句,第二阶段在715个NLLB翻译的三元组上微调,第三阶段使用英-基尼亚万达语翻译对对齐表示,第四阶段用2,936对高质量数据 refine。在SemRel2024-rw基准上,七检查点集成模型得分达0.7298,超越mE5-large 20.9%和OpenAI text-embedding-3-large 41.0%。
Current read
卢旺达团队发布了首个基尼亚万达语专用嵌入模型,四阶段训练效果显著,开源了模型和数据集。
11 sources58° AI heatLast updated 2026/8/27 10:42:48
Evidence
11 sourcesRelated source 1techcrunch
View sourceRelated source 2IT之家
View sourceRelated source 3The Rundown AI
View sourceRelated source 4Greg Brockman
View sourceRelated source 5Mustafa Suleyman
View sourceRelated source 6OpenAI
View sourceRelated source 7Decoder
View sourceRelated source 8@OpenAIDevs
View sourceRelated source 9Latent.Space
View sourceRelated source 10爱范儿
View sourceConflict check
Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.