BeamLLM combines computer vision with large language models to overcome key challenges in millimeter-wave communication systems, significantly reducing training overhead and latency.

• Extracts UE positional features from RGB images for improved beam prediction
• Aligns visual-temporal features with LLMs' semantic space through reprogramming
• Demonstrates how AI reasoning capabilities can enhance wireless communication performance
• Represents a novel cross-domain application of LLMs beyond traditional NLP tasks

This research addresses critical engineering challenges in next-generation wireless networks, promising faster, more reliable connections for applications requiring high bandwidth and low latency.

BeamLLM: Vision-Empowered mmWave Beam Prediction with Large Language Models