Scaling Pretrained Language Models in RAG Systems: Latency-Accuracy Trade-offs from 7B to 70B Parameters

This report synthesises findings from 4 peer-reviewed papers addressing the following research question: How does the trade-off between retrieval latency and generation accuracy in RAG systems vary when scaling pretrained language models from 7B to 70B parameters, as measured by response time and. 6 claims were extracted from source literature; 6 were independently verified against retrieved documents. An automated multi-reviewer quality assessment produced a score of 8.5/10. This report is a machine-generated literature synthesis and does not constitute original research.

Research goal: How does the trade-off between retrieval latency and generation accuracy in RAG systems vary when scaling pretrained language models from 7B to 70B parameters, as measured by response time and NaturalQuestions benchmark scores?

Autonomous literature synthesis. Automated review score: 8.5/10. Full text and citation available at Assignee Research.