AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

$G^2$-Reader: Dual Evolving Graphs for Multimodal Document QA

Yaxin Du, Junru Song, Yifan Zhou, Cheng Wang, Jiahao Gu, Zimeng Chen, Menglan Chen, Wen Yao, Yang Yang, Ying Wen, Siheng Chen

ICML 2026 · Vision-language and multimodal › Human and embodied modeling · approach: Multimodal retrieval-augmented modeling · 0 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 multimodal (visual) evidence to retrieval and generation (ablation)G2-Reader vanilla baseline · VisDoMBench 250-sample subsetaverage accuracy: 54.8 -> 63.2Improved
add Planning Graph on top of multimodal baseline (ablation)G2-Reader · VisDoMBench 250-sample subsetaverage accuracy: 63.2 -> 64.4Improved
add Content Graph on top of multimodal baseline (ablation)G2-Reader · VisDoMBench 250-sample subsetaverage accuracy: 63.2 -> 63.6Improved
combine Content Graph and Planning Graph (ablation)G2-Reader · VisDoMBench 250-sample subsetaverage accuracy: 63.2 -> 68.0Improved
increase Planning Graph refinement rounds from 0 to 3 (ablation)G2-Reader · VisDoMBench 250-sample subsetaverage accuracy: 64.8 -> 68.0Improved
increase refinement rounds beyond 3 (ablation)G2-Reader · VisDoMBench 250-sample subsetaverage accuracy: 68.0 -> 64.4Hurt
evolve Content Graph for 3 iterations instead of 0 (ablation)G2-Reader · VisDoMBench 250-sample subsetaverage accuracy: 61.8 -> 68.0Improved
evolve Content Graph for 5 iterations instead of 3 (ablation)G2-Reader · VisDoMBench 250-sample subsetaverage accuracy: 68.0 -> 63.6Hurt
replace VLM-based evolution with rule-based lite update (ablation)G2-Reader (T=3) · VisDoMBench 250-sample subsetaverage accuracy: 68.0 -> 67.0No clear effect
add structured subgraph readout (ablation)G2-Reader (T=3) · VisDoMBench 250-sample subsetaverage accuracy: 66.0 -> 68.0Improved
use G2-Reader with Qwen3-VL-32B instead of prompting the model directlyQwen3-VL-32B · VisDoMBench fullaverage accuracy: 29.90 -> 66.21Improved
use G2-Reader at k=5 versus vanilla RAG at k=20retrieval · recallrecall: 86.2 -> 86.3Improved

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