AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

TAdaRAG: Task Adaptive Retrieval-Augmented Generation via On-the-Fly Knowledge Graph Construction

Jie Zhang, Bo Tang, Wanzi Shao, Wenqiang Wei, Jihao Zhao, Jianqing Zhu, Zhiyu Li, Wen Xi, Zehao Lin, Feiyu Xiong, Yanchao Tan

AAAI 2026 · Large language models › Retrieval and knowledge · approach: RAG robustness and evaluation · 1 citations (OpenAlex)

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 prompt-based task-adaptive KG to NaiveRAG (ablation)Mistral-7B-Instruct · 2WikiMQAF1: 20.60 -> 38.48Improved
add supervised knowledge extraction fine-tuning (ablation)Mistral-7B-Instruct · UltraDomain LegalF1: 32.92 -> 39.32Improved
add REINFORCE task-adaptive KG construction (ablation)Mistral-7B-Instruct · UltraDomain LegalF1: 39.32 -> 49.88Improved
add REINFORCE stage (ablation)Qwen2.5-7B-Instruct · HotpotQAF1: 49.03 -> 49.23Improved
add REINFORCE stage (ablation)Qwen2.5-7B-Instruct · UltraDomain HealthF1: 41.35 -> 42.38Improved
use TAdaRAG instead of MemoRAGMistral-7B-Instruct · UltraDomain LegalF1: 51.20 -> 49.88Hurt
use TAdaRAG instead of MemoRAGQwen2.5-7B-Instruct · UltraDomain LegalF1: 47.60 -> 46.83Hurt
use TAdaRAG instead of MemoRAGMistral-7B-Instruct · 2WikiMQAF1: 30.30 -> 39.31Improved
use TAdaRAG instead of MemoRAGMistral-7B-Instruct · GovReportROUGE-L: 31.60 -> 36.41Improved
use TAdaRAG vs best baseline on NowNewsQA (human rating)Mistral-7B-Instruct · NowNewsQAoverall average human score: 7.720 -> 7.904Improved
increase parallel subgraphs from 3 to 4 or 5 (ablation)TAdaRAG 7B models · most datasetsF1Hurt
use 4 subgraphs instead of 3 (ablation)TAdaRAG · GovReportROUGE-LImproved

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