Fast and Accurate: Improving Personalized LLM Performance by Semantic Concentration Inspired Workload Split between RAG and Fine-Tuning

Name: Fast and Accurate: Improving Personalized LLM Performance by Semantic Concentration Inspired Workload Split between RAG and Fine-Tuning
Uploaded: 2026-10-18T19:23:35.035438+00:00
Duration: 700 s
Description: The proposed PEFT Union RAG (PUR) approach splits user data based on semantic concentration. It uses parameter-efficient fine-tuning for concentrated core knowledge and standard retrieval augmentation for dispersed data. This adaptive strategy improves personalized LLM accuracy and reduces inference latency by intelligently deciding whether to use fine-tuning or retrieval.

The proposed PEFT Union RAG (PUR) approach splits user data based on semantic concentration. It uses parameter-efficient fine-tuning for concentrated core knowledge and standard retrieval augmentation for dispersed data. This adaptive strategy improves personalized LLM accuracy and reduces inference latency by intelligently deciding whether to use fine-tuning or retrieval.

Open the full episode