pplx-embed-v2-late 技术报告
pplx-embed-v2-late Technical Report
Markus Krimmel · Louis Milliken · Ekaterina Glazkova · Maximilian Schall · Antoine Chaffin · Sedigheh Eslami · et al.
中文摘要
该报告介绍了 pplx-embed-v2-late：一对参数规模为 0.6B 和 9B 的多语言(multilingual)、多模态(multimodal)晚交互(late-interaction)模型。两者分别在对应规模级别中，于文本、图像和视觉文档检索基准上取得最先进结果。在视觉文档检索任务中，pplx-embed-v2-late-0.6b 的表现可与参数量高出数倍的模型竞争。模型通过从 18B 教师模型进行表示蒸馏训练，共享同一嵌入空间，支持非对称部署：文档可由 9B 模型建立索引，再由低延迟的 0.6B 模型处理查询。两种模型均已发布。
关键要点
- 010.6B 与 9B 模型覆盖文本、图像及视觉文档检索
- 02两者在各自规模级别的多组检索基准上达到最先进水平
- 030.6B 模型在视觉文档检索中可匹敌数倍规模模型
- 04表示蒸馏使两模型共享嵌入空间并支持非对称部署
- 05文档可用 9B 模型索引，由低延迟 0.6B 模型执行查询
解读
尚无解读。
原始英文摘要
arXiv:2610.29523v1 Announce Type: cross Abstract: In this report, we introduce pplx-embed-v2-late, a pair of 0.6B and 9B multilingual and multimodal late-interaction models. Both models achieve state-of-the-art results within their respective size classes across a range of text, image, and visual document retrieval benchmarks. On visual document retrieval tasks, pplx-embed-v2-late-0.6b is competitive with models several times larger than itself. The models are trained via representation distillation from an 18B teacher. Consequently, they share an embedding space that supports asymmetric deployment: documents can be indexed with the 9B model and queried with the low-latency 0.6B model. Both models are available at https://huggingface.co/collections/perplexity-ai/pplx-embed-v2.