LLama3 SatCom 70B: A Specialized LLM for Satellite Communications
Developed under the ESA ARTES programme as part of the SatcomLLM / SCEVA (SatCom Expert Virtual Assistant) project, LLama3 SatCom 70B is a high-capacity, instruction-tuned Large Language Model. Based on meta-llama/Llama-3.3-70B-Instruct, this 70 billion parameter model is purpose-built to support satellite communications (SatCom) professionals.
Key Capabilities & Differentiators
- SatCom Specialization: It is the largest and most capable LLM specifically fine-tuned for the SatCom domain, leveraging curated and synthetic SatCom QA datasets, including chain-of-thought annotated data.
- Enhanced Reasoning: Provides advanced reasoning, multi-step problem solving, and expert-level Q&A, with improved comprehension of technical systems and higher stability in mathematical reasoning compared to its 8B counterpart.
- Extended Context: Benefits from a 128k token context length, allowing it to process longer technical passages and integrate multiple data sources for complex analytical workflows.
- Performance: Achieves strong quantitative reasoning in link budgets and RF propagation tasks (0.695 accuracy on Math SatCom MCQA) and high factual precision in conceptual questions (0.947 accuracy on Open SatCom MCQA).
Intended Use Cases
- Advanced reasoning and Q&A for SatCom system design and analysis.
- Automated support for link budget calculations and RF engineering tasks.
- Conceptual guidance for 5G/6G NTN and inter-satellite network operations.
- Mission design evaluation and anomaly diagnosis assistance.
Limitations
- Does not access live telemetry or proprietary ESA mission data.
- Generated answers require validation by domain experts before operational use.
- Not suitable for safety-critical or real-time decision-making.