无需训练，推理时动态优化文档排序，显著提升检索效果。
Test-Time Training for Zero-Resource Dense Retrieval Reranking
- 用查询的前/后文档作伪正负样本，实时微调评分矩阵。
- 在6个基准上平均提升NDCG@10达2.1%，每查询增时不足10ms。
- 适合零资源场景，无需标注数据，跨领域泛化能力强。
密集检索器在候选生成阶段表现优异，但在零资源设置下缺乏有效的重排序能力。现有方法面临根本性困境：交叉编码器虽能提供强重排序性能，但需昂贵的监督训练且延迟高；而无监督的BM25重排序则在多数BEIR基准上持续降低密集检索性能。本文提出DART（测试时自适应重排序），通过在推理阶段动态调整打分函数解决此困境。对每个查询，将排名靠前的文档作为伪正例、靠后的作为伪负例，利用这些噪声但易得的监督信号，仅通过少量梯度更新即可适配双线性打分矩阵$W$。进一步引入置信度加权边界损失与跨查询动量缓存，实现查询间初始化加速。在六个BEIR基准上，DART相较密集检索基线平均提升相对NDCG@10达+2.1%，每查询额外延迟低于10ms，展现出强大的零样本性能增强与跨域泛化能力。
原文摘要 · Abstract (English)
Dense retrievers excel at first-stage candidate generation but lack effective reranking in zero-resource settings. Existing approaches face a fundamental dilemma: cross-encoders deliver strong reranking quality but require costly supervised training and incur high latency, while unsupervised BM25 reranking consistently degrades dense retrieval performance on most of BEIR benchmarks. We propose DART (Dense Adaptive Reranking at Test-time), which resolves this dilemma by adapting the scoring function at inference time. For each query, the top-ranked documents serve as pseudo-positive examples and the bottom-ranked as pseudo-negative examples, providing noisy but readily available supervision to adapt a bilinear scoring matrix $W$ via a small number of gradient updates. We further introduce a confidence-weighted margin loss and a cross-query momentum buffer that warm-starts adaptation across queries. On six BEIR benchmarks, DART achieves a mean per-dataset relative NDCG@10 gain of +2.1% over the dense retrieval baseline with under 10ms additional latency per query, demonstrating a powerful capability for zero-shot performance enhancement and cross-domain generalization.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。