Research map / Large language models / Retrieval and knowledge
SubGCache: Accelerating Graph-based RAG with Subgraph-level KV Cache
Qiuyu Zhu, Liang Zhang, Qianxiong Xu, Cheng Long, Jie Zhang
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 |
|---|---|---|---|
| add SubGCache subgraph-level KV cache | G-Retriever (Llama-2-7B) · Scene Graph | TTFT (ms): 938.44 -> 140.54 | Improved |
| add SubGCache | G-Retriever (Llama-3.2-3B) · OAG | TTFT (ms): 921.00 -> 141.19 | Improved |
| add SubGCache | G-Retriever (Llama-2-7B) · Scene Graph | ACC (%): 59.00 -> 66.00 | Improved |
| add SubGCache | GRAG (Mistral-7B) · Scene Graph | ACC (%): 57.00 -> 66.00 | Improved |
| add SubGCache | GRAG (Llama-3.2-3B) · OAG | ACC (%): 98.00 -> 97.00 | Hurt |
| add SubGCache | GRAG (Llama-3.2-3B) · OAG | TTFT (ms): 186.61 -> 124.44 | Improved |
| add SubGCache | G-Retriever (Mistral-7B) · Scene Graph and OAG | ACC (%): 66.00 -> 66.00 | No clear effect |
| add SubGCache | GRAG (Llama-3.2-3B) · OAG, 200 in-batch queries | ACC (%): 99.00 -> 98.50 | Hurt |
| increase in-batch query size from 50 to 150 | G-Retriever+SubGCache (Llama-3.2-3B) · Scene Graph | TTFT speedup: 3.40x -> 5.04x TTFT speedup | Improved |
| use fewer clusters (1 cluster) (ablation) | SubGCache · Scene Graph | TTFT and ACC | Improved |
| vary cluster number (more clusters) (ablation) | SubGCache · Scene Graph, OAG | ACC and TTFT | Mixed |
| clustering overhead (graph encoding, hierarchical clustering, merging) (ablation) | SubGCache · Scene Graph | latency share: <2.1% of total latency | No clear effect |
Closest papers by meaning
- Simple is Effective: The Roles of Graphs and Large Language Models in Knowledge-Graph-Based Retrieval-Augmented Generation (ICLR 2025 · similarity 0.91)
- Think-on-Graph 2.0: Deep and Faithful Large Language Model Reasoning with Knowledge-guided Retrieval Augmented Generation (ICLR 2025 · similarity 0.89)
- GFM-RAG: Graph Foundation Model for Retrieval Augmented Generation (NeurIPS 2025 · similarity 0.89)
- Inference Scaling for Long-Context Retrieval Augmented Generation (ICLR 2025 · similarity 0.89)
- A Systematic Exploration of Knowledge Graph Alignment with Large Language Models in Retrieval Augmented Generation (AAAI 2025 · similarity 0.89)
- Fast Think-on-Graph: Wider, Deeper and Faster Reasoning of Large Language Model on Knowledge Graph (AAAI 2025 · similarity 0.89)
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.
