Mistral-7B and DeBERTa Hybrid Improves Multi-Author Style Detection

Combining Mistral-7B embeddings fine-tuned via cross-domain transfer learning with DeBERTa embeddings improves authorship verification. This hybrid approach addresses the limitations of off-the-shelf LLMs in detecting stylistic changes across different user-generated text domains.