AI Paper Atlas

Research map / Graph learning / Knowledge and temporal graphs

Youtu-GraphRAG: Vertically Unified Agents for Graph Retrieval-Augmented Complex Reasoning

Junnan Dong, Siyu An, Yifei Yu, Qian-Wen Zhang, Linhao Luo, Xiao Huang, di yin, Yunsheng Wu, Xing Sun

ICLR 2026 · Graph learning › Knowledge and temporal graphs · approach: Logical query answering and rule reasoning

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 changedWhereMeasuredEffect
add iterative reasoning and reflection agent (ablation)Youtu-GraphRAG · 2Wiki (reject)top-20 accuracy: 57.80 -> 77.60Improved
add iterative reasoning and reflection agent (ablation)Youtu-GraphRAG · MuSiQue (reject)top-20 accuracy: 40.00 -> 47.50Improved
add iterative reasoning and reflection agent (ablation)Youtu-GraphRAG · HotpotQA (reject)top-20 accuracy: 75.30 -> 81.20Improved
add community detection (knowledge tree) (ablation)Youtu-GraphRAG · HotpotQA (reject)top-20 accuracy: 79.50 -> 81.20Improved
add community detection (knowledge tree) (ablation)Youtu-GraphRAG · 2Wiki (reject)top-20 accuracy: 75.10 -> 77.60Improved
add schema guidance for extraction and decomposition (ablation)Youtu-GraphRAG · AnonyRAG-CHS (open)top-20 accuracy: 35.61 -> 42.88Improved
add schema guidance for extraction and decomposition (ablation)Youtu-GraphRAG · AnonyRAG-ENG (open)top-20 accuracy: 40.32 -> 43.26Mixed
add schema guidance (ablation)Youtu-GraphRAG · MuSiQue (reject)top-20 accuracy: 45.60 -> 47.50Improved
use Youtu-GraphRAG without agent versus strongest baselinesYoutu-GraphRAG w/o Agent · HotpotQA, MuSiQue, G-Bench (open)top-20 accuracy: HotpotQA 81.80 (HippoRAG2) -> 83.70Improved
use Youtu-GraphRAG instead of graph baselines for graph construction token costYoutu-GraphRAG · six benchmarks (construction tokens)top-20 accuracy: up to 90.71% token savingImproved
use reject mode instead of open mode (disallow parametric knowledge)Naive RAG · 2Wikitop-20 accuracy: 70.3 -> 38.9Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy