AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Multimodal Table Understanding with Difficulty-aware Reinforcement Learning

Chaohu Liu, Haoyu Cao, YongXiang Hua, Linli Xu

AAAI 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Multimodal LLM benchmarks · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes a difficulty-aware reinforcement learning approach for multimodal table understanding that accounts for table structural complexity (merged-cell ratio, total cell count).

Problem

Multimodal table understanding, which integrates cell text, tabular structure, and visual presentation, is challenging because structural complexity (e.g., ratio of merged cells, total cell count) impedes existing models.

Approach

Quantifies table difficulty via intrinsic structural properties (merged-cell ratio and total number of cells) and uses difficulty-aware reinforcement learning to train on tables of varying complexity. Details beyond the abstract are not stated here.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add Stage 2 RL after Stage 1 perception training (vs Stage 1 only) (ablation)MM-Table-R1-3B · HiTabaccuracy: 66.83 -> 76.14Improved
add Stage 1 table reconstruction RL before Stage 2 on TabFact (ablation)MM-Table-R1-3B · TabFactaccuracy: 83.53 -> 87.13Improved
add Stage 1 table reconstruction RL before Stage 2 on ToTTo (ablation)MM-Table-R1-3B · ToTToaccuracy: 55.32 -> 78.67Improved
add Stage 1 table reconstruction RL before Stage 2 on TableVQA (ablation)MM-Table-R1-3B · TableVQAaccuracy: 72.06 -> 78.60Improved
add Stage 1 table reconstruction RL before Stage 2 on HiTab (ablation)MM-Table-R1-3B · HiTabaccuracy: 71.21 -> 76.14Improved
use RL Stage 1 instead of SFT for perception (ablation)MM-Table-R1-3B · TableVQAaccuracy: 73.93 -> 78.60Improved
use RL Stage 1 instead of SFT for perception (ablation)MM-Table-R1-3B · ToTToaccuracy: 63.61 -> 78.67Improved
use RL Stage 1 instead of SFT for perception (ablation)MM-Table-R1-3B · TabFactaccuracy: 82.19 -> 87.13Improved
SFT training on Qwen2.5-VL-3B baseQwen2.5-VL-3B · TableVQAaccuracy: 63.94 -> 62.73Hurt
SFT training on Qwen2.5-VL-3B baseQwen2.5-VL-3B · HiTabaccuracy: 63.42 -> 67.91Improved
Stage 1 only on base modelQwen2.5-VL-3B · ToTToaccuracy: 48.61 -> 78.71Improved
Stage 2 only on base modelQwen2.5-VL-3B · TableVQAaccuracy: 63.94 -> 72.06Improved

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