jevIMAGE
发票流水线：Jev 分类 vs 4B 微调模型决策
收件箱分类 Jev 200/200 全对无需训练；但判断是否付款时，4B 微调模型 0.98 胜过 Jev 的 0.79。
Jacek Golebiowski
@j_golebiowski2026.09.24收录原推发布于 2026.09.23
Jacek Golebiowski @j_golebiowski原推发布于 2026.09.23
Where does Jev end and a fine-tuned small model begin? We built an invoice pipeline with both; sorting the inbox: Jev 200/200, nothing to train. Deciding whether to pay: Jev lands at 0.79 but a 4B fine-tuned to reason first gets 0.98. Details from @distil_labs below.
x.com/j_golebiowski/status/2102782627442819504采集于 2026.09.24 10:15