Standalone model produced by merging the LoRA adapter Dospacite/xai-phishing-deepseek-r1-qwen-7b into deepseek-ai/DeepSeek-R1-Distill-Qwen-7B.

This repository contains complete model weights and does not require PEFT or the adapter repository at inference time. It expects the structured webpage feature prompt documented in the original adapter repository.

The model classifies structured webpage evidence as phishing or benign and returns an explanation. Its output is decision support and should be reviewed alongside the cited webpage features.