Researchers have developed an offline language-based support system called MATE (Mission Assistant for Training and Evaluation) to aid air combat decision-making and pilot training. This system utilizes retrieval-augmented generation (RAG) to answer natural language queries using information from tactical manuals, operating entirely with open-source components for local functionality. Evaluations by three Brazilian Air Force pilots showed high satisfaction with the system's response time and answer quality, with Mistral NeMo 12B performing best among the tested models (Llama 3.1 8B, Qwen 2.5 14B). The pilots identified the relevance of retrieved content as an area for improvement. AI
IMPACT Demonstrates potential for offline, localized AI decision support in specialized domains like military operations.
RANK_REASON Academic paper detailing a new system and its evaluation. [lever_c_demoted from research: ic=1 ai=1.0]
- arXiv
- Brazilian Air Force
- Hugging Face
- Llama-3.1:8b
- MATE
- Mistral Nemo 12B
- Qwen-2.5 14B
- retrieval-augmented generation
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →