Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2509.16889 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915891398049792 |
|---|---|
| author | Kang, Xiaoqiang Wu, Shengen Wang, Zimu Liu, Yilin Jin, Xiaobo Huang, Kaizhu Wang, Wei Yue, Yutao Huang, Xiaowei Wang, Qiufeng |
| author_facet | Kang, Xiaoqiang Wu, Shengen Wang, Zimu Liu, Yilin Jin, Xiaobo Huang, Kaizhu Wang, Wei Yue, Yutao Huang, Xiaowei Wang, Qiufeng |
| contents | Existing table understanding methods face challenges due to complex table structures and intricate logical reasoning. While supervised finetuning (SFT) dominates existing research, reinforcement learning (RL), such as Group Relative Policy Optimization (GRPO), has shown promise but struggled with low initial policy accuracy and coarse rewards in tabular contexts. In this paper, we introduce Table-R1, a three-stage RL framework that enhances multimodal table understanding through: (1) Warm-up that prompts initial perception and reasoning capabilities, (2) Perception Alignment GRPO (PA-GRPO), which employs continuous Tree-Edit-Distance Similarity (TEDS) rewards for recognizing table structures and contents, and (3) Hint-Completion GRPO (HC-GRPO), which utilizes fine-grained rewards of residual steps based on the hint-guided question. Extensive experiments demonstrate that Table-R1 can boost the model's table reasoning performance obviously on both held-in and held-out datasets, outperforming SFT and GRPO largely. Notably, Qwen2-VL-7B with Table-R1 surpasses larger specific table understanding models (e.g., Table-LLaVA 13B), even achieving comparable performance to the closed-source model GPT-4o on held-in datasets, demonstrating the efficacy of each stage of Table-R1 in overcoming initialization bottlenecks and reward sparsity, thereby advancing robust multimodal table understanding. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_16889 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Can GRPO Boost Complex Multimodal Table Understanding? Kang, Xiaoqiang Wu, Shengen Wang, Zimu Liu, Yilin Jin, Xiaobo Huang, Kaizhu Wang, Wei Yue, Yutao Huang, Xiaowei Wang, Qiufeng Computation and Language Existing table understanding methods face challenges due to complex table structures and intricate logical reasoning. While supervised finetuning (SFT) dominates existing research, reinforcement learning (RL), such as Group Relative Policy Optimization (GRPO), has shown promise but struggled with low initial policy accuracy and coarse rewards in tabular contexts. In this paper, we introduce Table-R1, a three-stage RL framework that enhances multimodal table understanding through: (1) Warm-up that prompts initial perception and reasoning capabilities, (2) Perception Alignment GRPO (PA-GRPO), which employs continuous Tree-Edit-Distance Similarity (TEDS) rewards for recognizing table structures and contents, and (3) Hint-Completion GRPO (HC-GRPO), which utilizes fine-grained rewards of residual steps based on the hint-guided question. Extensive experiments demonstrate that Table-R1 can boost the model's table reasoning performance obviously on both held-in and held-out datasets, outperforming SFT and GRPO largely. Notably, Qwen2-VL-7B with Table-R1 surpasses larger specific table understanding models (e.g., Table-LLaVA 13B), even achieving comparable performance to the closed-source model GPT-4o on held-in datasets, demonstrating the efficacy of each stage of Table-R1 in overcoming initialization bottlenecks and reward sparsity, thereby advancing robust multimodal table understanding. |
| title | Can GRPO Boost Complex Multimodal Table Understanding? |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2509.16889 |