Jerry Liu 比较 Jev 与开源模型的文档任务表现
系统一模型对于各种需要快速决策的文档任务非常有用:
- 方向检测
- 语言检测
- 分类
- 分割
我们在多种此类任务上,将 Jev 与其他开源模型进行了比较,其中有通用分类器,也有文档专用模型。我们测量了准确率、成本和延迟。
在这里的大多数基准测试中,Jev 在这三个维度上都名列前茅。我们很期待使用它,并从中学习,继续推进文档理解能力的前沿!
YouTube:https://www.youtube.com/watch?v=MLZ1dUOZG74 代码仓库:https://github.com/run-llama/jev_vs_oss
对照原文
System one models are extremely useful for a variety of document tasks that require fast decisions: * orientation detection * language detection * classification * splitting We benchmarked Jev with other OSS models (some general classifiers, some document-specific) against a variety of these tasks. We measured accuracy, cost, and latency. Jev tops most of the benchmarks here across the 3 dimensions. We're excited to use it and learn from it to continue advancing the frontiers of doc understanding! YT: https://t.co/Xr3eKX9oF4 Repo: https://t.co/Ma77kEjMFV