Researchers have introduced DrIG, a novel framework for universal multimodal retrieval that utilizes dual-role identifiers to enhance generative information retrieval. This approach addresses limitations in existing methods, such as prefix-level errors and the predominantly unimodal nature of current systems. DrIG supports diverse retrieval tasks across text, image, and mixed image-text data by assigning each candidate a single identifier that functions both sequentially for autoregressive decoding and as a set for prefix-independent relevance priors. Experiments on the M-BEIR benchmark and text-to-image datasets demonstrate that DrIG outperforms state-of-the-art generative multimodal baselines, offering a favorable efficiency-effectiveness trade-off. AI
IMPACT This framework could improve the efficiency and accuracy of information retrieval systems across various data types.
RANK_REASON The cluster contains a research paper detailing a new framework for generative multimodal retrieval. [lever_c_demoted from research: ic=1 ai=1.0]
Read on arXiv cs.IR (Information Retrieval) →
- CatalyzeX Code Finder for Papers
- Connected Papers
- DagsHub
- Generative Information Retrieval
- Generative Universal Multimodal Retrieval with Dual-role Identifiers
- Gotit.pub
- Hugging Face
- Litmaps
- M-BEIR
- ScienceCast
- scite Smart Citations
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
