Research map / Vision-language and multimodal / Human and embodied modeling
Beyond Text-Only: Towards Multimodal Table Retrieval in Open-World
Da Li, Keping Bi, Jiafeng Guo, Wei Yuan, Fan Yang, Tingting Gao, Xueqi Cheng
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 |
|---|---|---|---|
| represent table content as image instead of OCR text (ablation) | Qwen2-VL-2B retriever · TaR-ViR | NDCG@5: 65.91 -> 67.42 | Improved |
| represent table title as image instead of text (ablation) | Qwen2-VL-2B retriever · TaR-ViR | Recall@10: 91.69 -> 82.30 | Hurt |
| use original web text instead of OCR text (ablation) | Qwen2-VL-2B text retriever · TaR-ViR | Recall@10 vs NDCG@5: R@10 89.00 -> 90.30; N@5 65.91 -> 65.20 | Mixed |
| use image content instead of OCR text on Hard tables (ablation) | Qwen2-VL-2B retriever · TaR-ViR Hard tables | NDCG@5: 62.82 -> 65.29 | Improved |
| use image content instead of OCR text on Easy tables (ablation) | Qwen2-VL-2B retriever · TaR-ViR Easy tables | NDCG@5: 71.86 -> 68.85 | Hurt |
| scale VLM2Vec from 2.21B to 7.07B parameters (ablation) | VLM2Vec · TaR-ViR | NDCG@5: 67.89 -> 70.56 | Improved |
| use larger text embedders (Qwen3-Embedding 4.05B) instead of BGE 109M (ablation) | text retriever · TaR-ViR | NDCG@5: 72.82 -> 71.93 | Hurt |
| use multimodal instead of text retriever for RAG with Qwen3-8B | RAG pipeline · TaR-ViR QA | accuracy: 58.38 -> 58.55 | No clear effect |
| feed table as image instead of text to Qwen2.5-VL-7B generator | RAG pipeline · TaR-ViR QA | accuracy: 38.51 -> 40.87 | Improved |
Closest papers by meaning
- TableDART: Dynamic Adaptive Multi-Modal Routing for Table Understanding (ICLR 2026 · similarity 0.89)
- TabPedia: Towards Comprehensive Visual Table Understanding with Concept Synergy (NeurIPS 2024 · similarity 0.87)
- TABLET: A Large-Scale Dataset for Robust Visual Table Understanding (ICLR 2026 · similarity 0.86)
- Multimodal Tabular Reasoning with Privileged Structured Information (NeurIPS 2025 · similarity 0.86)
- NeedleInATable: Exploring Long-Context Capability of Large Language Models towards Long-Structured Tables (NeurIPS 2025 · similarity 0.85)
- SynTab-LLaVA: Enhancing Multimodal Table Understanding with Decoupled Synthesis (CVPR 2025 · similarity 0.85)
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.
