FinVector-Market-4B：面向结构化金融任务的 LoRA 适配受控研究

arXiv cs.LG 论文
lora fine-tuning finance structured-output qwen benchmark json-schema

摘要

本文提出 FinVector-Market-4B，针对 Qwen3.5-4B 在 22,000 条语料上开展结构化金融任务的 LoRA（rank-16）适配研究。研究表明，在提示契约（prompt contract）一致的条件下，紧凑的领域适配能够带来超出单纯输出格式学习的显著任务收益。

arXiv:2610.08882v1 Announce Type: new Abstract: FinVector-Market-4B adapts Qwen/Qwen3.5-4B with rank-16 LoRA on a 22,000-example corpus for structured financial tasks. We evaluate the base and adapted models on the same 600-example benchmark under implicit and explicit JSON-schema contracts. Supplying the schema alone raises base-model JSON validity from 0% to 91.3%. Under matched explicit prompting, the frozen scores improve from 14.7% to 40.0% for FinQA answer exact match, from 48.0% to 82.7% for calculator-expression correctness, from 20.1% to 89.5% for scenario branch-label agreement, and from 52.4% to 87.2% for implication-direction agreement. A post-hoc policy-scoring audit shows that the reported macro-F1 decline reflects a changing label set; using the same three target classes gives 77.4% for the base and 83.1% for the adapter. Filing overlap and calculator-target inconsistencies qualify the benchmark's generalization claims. The results show that compact financial domain adaptation can produce substantial task-specific gains beyond output-format learning under matched prompting, with gains bounded by the evaluated task distribution and prompt contract.
查看原文
查看缓存全文

缓存时间: 2026/10/08 12:49 

# FinVector-Market-4B: A Controlled Study of LoRA Adaptation for Structured Financial Tasks
Source: [https://arxiv.org/abs/2610.08882](https://arxiv.org/abs/2610.08882)
[View PDF](https://arxiv.org/pdf/2610.08882)

> Abstract:FinVector\-Market\-4B adapts Qwen/Qwen3\.5\-4B with rank\-16 LoRA on a 22,000\-example corpus for structured financial tasks\. We evaluate the base and adapted models on the same 600\-example benchmark under implicit and explicit JSON\-schema contracts\. Supplying the schema alone raises base\-model JSON validity from 0% to 91\.3%\. Under matched explicit prompting, the frozen scores improve from 14\.7% to 40\.0% for FinQA answer exact match, from 48\.0% to 82\.7% for calculator\-expression correctness, from 20\.1% to 89\.5% for scenario branch\-label agreement, and from 52\.4% to 87\.2% for implication\-direction agreement\. A post\-hoc policy\-scoring audit shows that the reported macro\-F1 decline reflects a changing label set; using the same three target classes gives 77\.4% for the base and 83\.1% for the adapter\. Filing overlap and calculator\-target inconsistencies qualify the benchmark's generalization claims\. The results show that compact financial domain adaptation can produce substantial task\-specific gains beyond output\-format learning under matched prompting, with gains bounded by the evaluated task distribution and prompt contract\.

## Submission history

From: Alina Khaybullina \[[view email](https://arxiv.org/show-email/8049f920/2610.08882)\] **\[v1\]**Tue, 6 Oct 2026 10:46:58 UTC \(567 KB\)

相似文章

可以多小？LoRA微调270M-8B模型用于金融交易中的商户信息提取

arXiv cs.AI

本文是一项面向部署的研究，比较了24种模型变体（参数规模从270M到8B）在金融交易字符串中提取商户信息时的LoRA微调效果。作者发现，像Qwen 3.5 4B这样的小型模型可达到96.6%的F1分数，仅比8B基线低0.35个百分点，同时显著降低了延迟和成本。