LLM分类器为何难用却难弃
推荐指数 63.0 NO. 011 · 2026.09.18
发布2026/09/17Score79Comments16
为什么值得看
文章剖析了将LLM直接用作分类器的核心缺陷：输出校准差、阈值控制困难、难以融合结构化数据。这些问题导致LLM分类器在工程落地时精度与可控性不足，尽管表面效果尚可。
编辑判断
很多团队把LLM当分类器用，是因为省了标注和训练模型的成本，但往往在A/B测试阶段才发现校准问题。更务实的做法是把LLM当作特征提取器——让它输出embedding或中间推理，再喂给传统的逻辑回归或XGBoost，这样既能利用LLM的理解能力，又能保留阈值调节和可解释性。
已经有团队在实践中验证了这个路径：用LLM生成结构化特征后，小模型分类的F1反而比端到端LLM高5-10个百分点，延迟还降了一个数量级。如果你现在的系统里LLM分类器的误杀率居高不下，建议优先尝试这个拆分架构。
社区反馈
意见分歧 16 条评论
核心争论：LLM应直接做分类器还是仅作特征提取器供传统模型决策
I don't understand the point they are trying to make. It's very often (always?) the case that something general also solves particular problems. A sorting algorithm is an implementation of min() A parser also is a syntax checker. A route planner is a reachability checker. A compu
I took the point as: don't make the LLM the classifier. Use it to turn messy input into useful features, then let a normal model make the actual decision. That gives you thresholds/calibration you can inspect. What I'm not sure about is how stable those features are when you switch the underlyi
Much like how you shouldn't ask the LLM to solve a (repeated, logical) problem, but you should instead prompt it to generate code that you can inspect/test/fix/reuse.