G-Refer

Pairwise ranking beats RL for LLM explanation selection in recommendation systems

Researchers have developed a new method for selecting explanations from large language models (LLMs) in recommendation systems, significantly reducing serving costs and latency. By pre-generating a pool of explanations …