OCR-Agent: Agentic OCR with Capability and Memory Reflection
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866908850739740672 |
|---|---|
| author | Wen, Shimin Zhang, Zeyu Bian, Xingdou Zhu, Hongjie He, Lulu Shama, Layi Ergu, Daji Cai, Ying |
| author_facet | Wen, Shimin Zhang, Zeyu Bian, Xingdou Zhu, Hongjie He, Lulu Shama, Layi Ergu, Daji Cai, Ying |
| contents | Large Vision-Language Models (VLMs) have demonstrated significant potential on complex visual understanding tasks through iterative optimization methods.However, these models generally lack effective self-correction mechanisms, making it difficult for them to independently rectify cognitive biases. Consequently, during multi-turn revisions, they often fall into repetitive and ineffective attempts, failing to achieve stable improvements in answer quality.To address this issue, we propose a novel iterative self-correction framework that endows models with two key capabilities: Capability Reflection and Memory Reflection. This framework guides the model to first diagnose errors and generate a correction plan via Capability Reflection, then leverage Memory Reflection to review past attempts to avoid repetition and explore new solutions, and finally, optimize the answer through rigorous re-reasoning. Experiments on the challenging OCRBench v2 benchmark show that OCR-Agent outperforms the current open-source SOTA model InternVL3-8B by +2.0 on English and +1.2 on Chinese subsets, while achieving state-of-the-art results in Visual Understanding (79.9) and Reasoning (66.5) - surpassing even larger fine-tuned models. Our method demonstrates that structured, self-aware reflection can significantly enhance VLMs' reasoning robustness without additional training. Code: https://github.com/AIGeeksGroup/OCR-Agent. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_21053 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | OCR-Agent: Agentic OCR with Capability and Memory Reflection Wen, Shimin Zhang, Zeyu Bian, Xingdou Zhu, Hongjie He, Lulu Shama, Layi Ergu, Daji Cai, Ying Computer Vision and Pattern Recognition Large Vision-Language Models (VLMs) have demonstrated significant potential on complex visual understanding tasks through iterative optimization methods.However, these models generally lack effective self-correction mechanisms, making it difficult for them to independently rectify cognitive biases. Consequently, during multi-turn revisions, they often fall into repetitive and ineffective attempts, failing to achieve stable improvements in answer quality.To address this issue, we propose a novel iterative self-correction framework that endows models with two key capabilities: Capability Reflection and Memory Reflection. This framework guides the model to first diagnose errors and generate a correction plan via Capability Reflection, then leverage Memory Reflection to review past attempts to avoid repetition and explore new solutions, and finally, optimize the answer through rigorous re-reasoning. Experiments on the challenging OCRBench v2 benchmark show that OCR-Agent outperforms the current open-source SOTA model InternVL3-8B by +2.0 on English and +1.2 on Chinese subsets, while achieving state-of-the-art results in Visual Understanding (79.9) and Reasoning (66.5) - surpassing even larger fine-tuned models. Our method demonstrates that structured, self-aware reflection can significantly enhance VLMs' reasoning robustness without additional training. Code: https://github.com/AIGeeksGroup/OCR-Agent. |
| title | OCR-Agent: Agentic OCR with Capability and Memory Reflection |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2602.21053 |