Researchers have developed a method for customizing generative AI agents for specialized fields like transportation engineering. By using a curated dataset of U.S. transportation documents, they fine-tuned six large language models (LLMs) with a low-rank adaptation (LoRA) framework. The study found that Qwen2.5-7B and LLaMA-3.1-8B models showed the best performance in understanding technical content and reasoning within the domain, as measured by BLEU-4 and ROUGE scores. This approach offers a reproducible way to create domain-specific AI agents for applications in research, design, planning, and policy. AI
IMPACT This research provides a framework for adapting LLMs to specialized domains, potentially improving AI applications in fields like transportation engineering.
RANK_REASON Academic paper detailing a new methodology for domain-specific LLM adaptation. [lever_c_demoted from research: ic=1 ai=1.0]
- arXiv
- BLEU-4
- generative artificial intelligence
- large-language models
- Llama-3.1:8b
- Lora
- qwen2.5:7b
- Rouge
- transportation engineering
- U.S.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
