Original Summary

I’ve been experimenting with a local approach to some of the classification tasks people are using Jev for. This approach uses text embeddings + logistic regression<p>On Banking77, which contains 77 categories of banking support questions, I get 94.25% using bge-large-en-v1.5 for embeddings, and 93.28% with all-MiniLM-L6-v2 (only the classifier gets trained, the embeddings model stays unchanged). For comparison:<p><pre><code> Model Accuracy Size&#x2F;training time - IntenDD (SOTA): 94.86% (~350M params, hours on GPU) - This script: 94.25% (642 KB classifier, 3s on CPU) - ModernBERT fine-tuned: 93.99% (149M params, minutes on GPU) - BERT fine-tuned: 92.76% (110M params, minutes on GPU) - Jev (zero-shot): 87.0% (closed API) - Laya (zero-shot): 42.5% (421M params) </code></pre> MiniLM’s encoder is roughly 91 MB in FP32, BGE is about 1.2 GB, the classifier and scaler are only 642 KB<p>The linked gist evaluates all 3,080 official test examples. Running python banking77_gist.py uses MiniLM by default, and you can choose the larger BGE encoder. Everything runs locally (although it needs to download the embedding model on the first run)<p>This is just a proof of concept, but it shows you can get very good results without needing GPUs, large amounts of data or hours of training, if you mostly just need classification<p>The direction I&#x27;m now exploring, is a system that handles familiar decisions with classifiers first, but asks an LLM for help when needed, then uses the LLM responses to automatically improve its local classifiers<p>What are you using Jev&#x2F;Laya for?

  • 情报分类：技术学习与提效
  • 分类依据：内容涉及技术、AI、软件工具或工程实践
  • 信息来源：Hacker News 新项目
  • 发布时间：2026/9/27 01:41:51