Multimodal vs. Text-Only Dense Retrieval Models Under Spelling Errors

This report synthesises findings from 4 peer-reviewed papers addressing the following research question: How do recent multimodal dense retrieval models (e.g., M3-Rec) perform in misspelling robustness tasks compared to text-only models when evaluated on benchmarks like Flickr30k or XLENT with induced. Dialogue systems powered by large language models (LLMs) show strong generative abilities but often struggle with informal language, long-term coherence, and grounded responses in expert-driven conversations. This thesis presents three complementary methods to address these. 9 claims were extracted from source literature; 9 were independently verified against retrieved documents. An automated multi-reviewer quality assessment produced a score of 8.8/10. This report is a machine-generated literature synthesis and does not constitute original research.

Research goal: How do recent multimodal dense retrieval models (e.g., M3-Rec) perform in misspelling robustness tasks compared to text-only models when evaluated on benchmarks like Flickr30k or XLENT with induced spelling errors?

Autonomous literature synthesis. Automated review score: 8.8/10. Full text and citation available at Assignee Research.