STAR: Semantic-Tuned and Tail-Adaptive Retriever for Graph-Augmented Generation
The article introduces STAR, a new retriever designed to enhance Graph Retrieval Augmented Generation for multi-hop question answering. STAR addresses the challenges of semantic biases in existing methods by employing token-level interaction and path-weighted contrastive learning. Experimental results show that STAR outperforms previous models, improving retrieval and question-answering performance across benchmark datasets.
- ▪STAR is a semantic-tuned and tail-adaptive retriever for Graph Retrieval Augmented Generation.
- ▪It mitigates Semantic Shortcut Bias and Long-Tail Path Bias through innovative learning paradigms.
- ▪Extensive experiments indicate that STAR achieves average retrieval performance gains of 1.8% and LLM QA performance improvements of 2.2%.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Information Retrieval arXiv:2605.18765 (cs) [Submitted on 11 Apr 2026] Title:STAR: Semantic-Tuned and Tail-Adaptive Retriever for Graph-Augmented Generation Authors:Shuai Li, Chen Huang, Duanyu Feng, Wenqiang Lei, See-Kiong Ng View a PDF of the paper titled STAR: Semantic-Tuned and Tail-Adaptive Retriever for Graph-Augmented Generation, by Shuai Li and 4 other authors View PDF HTML (experimental) Abstract:To augment Large Language Models (LLMs) for multi-hop question answering, a mainstream solution within Graph Retrieval Augmented Generation (GraphRAG) leverages lightweight retrievers to efficiently extract information from a given Knowledge Graph (KG). However, existing methods often overlook the inherent challenge of sparse semantic information in graphs.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
