arXiv · cs.AI · Peichun Hua, Yunming Xiao · 11 小时前模型评分39
Matryoshka Hash Representations（MHR）：面向模型感知紧凑语义检索的哈希表示
Matryoshka Hash Representations for Model-Aware Compact Semantic Retrieval
导读
研究者提出 Matryoshka Hash Representations（MHR），一种两阶段方法：先学习较长的二进制码，再冻结模型并训练零初始化的残差码适配器，使短前缀可直接搜索。
分类置信度
中 · 待人工复核
- 存在未解决的相邻类别：ai-tooling，需要人工复核。
- 只有单一媒体来源，需要人工复核。
来源级别：媒体与个人
事件类型：研究发布
证据摘录
- “We introduce Matryoshka Hash Representations (MHR), a two-stage procedure that separates full-width training from prefix organization.”
- “Documents are stored at one bit per coordinate, queries keep continuous logits as in PQ, and search runs on FAISS FastScan.”
- “Trained on MS MARCO and zero-shot transferred to seven BEIR datasets, one MHR code reaches 0.502 NDCG@10 at 16 bytes and 0.556 at 32 bytes, against 0.404 and 0.524 for the best compact baselines”
来源：arXiv · cs.AI · arxiv.org