English(EN) Generative Retrieval via Diffusion Transformer with Metric-Ordered Sequence Training and Hybrid-Policy Preference Optimization

新的扩散 Transformer 框架增强了模式保持属性检索

作者 PulseAugur 编辑部 · [3 个来源] · 2026-06-25 11:29

研究人员引入了一种名为 MO-DiT+HPPO 的新颖框架，用于模式保持属性检索。该方法使用扩散 Transformer 生成满足特定属性并保持给定模式的查询嵌入，解决了传统基于嵌入的检索的局限性。该框架采用分阶段训练，包括度量排序序列训练和混合策略偏好优化，以提高跨各种领域的检索准确性。 AI

影响这项研究可能导致更复杂的检索系统，能够理解和维护数据中的复杂模式。

排序理由该集群描述了一篇详细介绍新 AI 模型和训练框架的研究论文。

AI 生成摘要 · Google Gemini · 来自 3 个来源。我们如何撰写摘要 →

报道来源 [3]

arXiv cs.AI TIER_1 English(EN) · Chenghao Liu, Yu Zhang, Zhongtao Jiang, Kun Xu, Zhenwei An, Renzhi Wang, Zhao Wang, Jiachen Zhang, Yuxiao Zhang, Kun Xu, Songfang Huang · 2026-06-26 04:00

基于度量排序序列训练和混合策略偏好优化的扩散Transformer生成式检索

arXiv:2606.26899v1 Announce Type: new Abstract: Embedding-based retrieval ranks items by their similarity to a query in a shared vector space and usually aims to return the highest-scoring items. In many production settings this is not what is wanted: given a seed set that expres…
Hugging Face Daily Papers TIER_1 English(EN) · 2026-06-25 11:29

基于度量排序序列训练和混合策略偏好优化的扩散 Transformer 生成式检索

Embedding-based retrieval ranks items by their similarity to a query in a shared vector space and usually aims to return the highest-scoring items. In many production settings this is not what is wanted: given a seed set that expresses a fine-grained pattern, one needs more items…
arXiv cs.AI TIER_1 English(EN) · Songfang Huang · 2026-06-25 11:29

基于度量排序序列训练和混合策略偏好优化的扩散Transformer生成式检索

Embedding-based retrieval ranks items by their similarity to a query in a shared vector space and usually aims to return the highest-scoring items. In many production settings this is not what is wanted: given a seed set that expresses a fine-grained pattern, one needs more items…