How Small Can You Go? LoRA Fine-Tuning 270M-8B Models for Merchant Information Extraction in Financial Transactions

In three linesComparative study of 24 model variants (270M-8B) fine-tuned with LoRA for merchant information extraction in financial transactions. Qwen 3.5 4B achieves 96.60% F1 (vs 96.95% for LLaMA 3.1-8B) with half the parameters. Qwen 3.5 0.8B reaches 94.75% F1, matching models 2.5-4x larger.

Summary generated by Claude — human-verified