Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
ZeroGR: A Generalizable and Scalable Framework for Zero-Shot Generative Retrieval
Weiwei Sun, Keyi Kong, Xinyu Ma, Shuaiqiang Wang, Dawei Yin, Maarten de Rijke, Zhaochun Ren, Yiming Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| increase number of instruction-tuning tasks (MS MARCO to OpenInstIR) (ablation) | ZeroGR · MAIR unseen subset | Acc@1 | Improved |
| increase generated pseudo-queries per document to eight (ablation) | ZeroGR · MAIR dev | Acc@1 | Improved |
| increase generated pseudo-queries per document to sixteen (ablation) | ZeroGR · MAIR dev | Acc@1 | Improved |
| increase backbone model size (Qwen2.5 series) (ablation) | ZeroGR · MAIR unseen tasks | Acc@1 | Improved |
| use reverse annealing decoding instead of greedy decoding (ablation) | ZeroGR · MAIR dev | Acc@1 and Recall@100 | Mixed |
| use reverse annealing decoding instead of nucleus sampling (ablation) | ZeroGR · MAIR dev | Acc@1 | Improved |
| use unified generated docid instead of RQ-VAE, summary, query or other docid designs (ablation) | ZeroGR · MAIR dev | Acc@1 | Improved |
| use a docid prefix of eight words (ablation) | ZeroGR docid · docid conflict rate | conflict rate: 0.45% | Improved |
| train query generator on diverse tasks instead of MS MARCO only (ablation) | ZeroGR · MAIR unseen subset | docid conflict rate | Improved |
Closest papers by meaning
- UniGen: A Unified Generative Framework for Retrieval and Question Answering with Large Language Models (AAAI 2024 · similarity 0.90)
- Generative Retrieval Meets Multi-Graded Relevance (NeurIPS 2024 · similarity 0.89)
- Learning to Rank in Generative Retrieval (AAAI 2024 · similarity 0.89)
- Retrieval is Accurate Generation (ICLR 2024 · similarity 0.89)
- Attention in Large Language Models Yields Efficient Zero-Shot Re-Rankers (ICLR 2025 · similarity 0.88)
- DOGR: Leveraging Document-Oriented Contrastive Learning in Generative Retrieval (AAAI 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
