MA-DPR Robustness and Inference Overhead in Out-of-Distribution Retrieval

This report synthesises findings from 14 peer-reviewed papers addressing the following research question: Does the robustness of MA-DPR in out-of-distribution retrieval scenarios justify the increased inference overhead compared to quantized Euclidean DPR models on edge AI accelerators. We propose a simple and effective re-ranking method for improving passage retrieval in open question answering. The re-ranker re-scores retrieved passages with a zero-shot question generation model, which uses a pre-trained language model to compute the probability of the input. 9 claims were extracted from source literature; 9 were independently verified against retrieved documents. An automated multi-reviewer quality assessment produced a score of 8.6/10. This report is a machine-generated literature synthesis and does not constitute original research.

Research goal: Does the robustness of MA-DPR in out-of-distribution retrieval scenarios justify the increased inference overhead compared to quantized Euclidean DPR models on edge AI accelerators?

Autonomous literature synthesis. Automated review score: 8.6/10. Full text and citation available at Assignee Research.