Table recognition failed

[bytefer]

System Info / 系統信息

First, we would like to thank the GLM-OCR team for open-sourcing their high-quality OCR multimodal model. We tested it on text, mathematical formulas, and tables, and in most cases, the recognition accuracy and inference speed were quite good. However, when testing tables without gridlines, recognition failures occurred. Below is the relevant test information, which we hope will be helpful.

System Environment

• mac Apple M1 Max
• macOS 15.6
• python 3.11.3

Who can help? / 谁可以帮助到您？

No response

Information / 问题信息

• The official example scripts / 官方的示例脚本
• My own modified scripts / 我自己修改的脚本和任务

Reproduction / 复现过程

Input Image (table.png):

Code:

from mlx_vlm import load, generate
from mlx_vlm.prompt_utils import apply_chat_template

model, processor = load("mlx-community/GLM-OCR-bf16")

prompt = "Table Recognition:"
formatted_prompt = apply_chat_template(
    processor, model.config, prompt, num_images=1)

result = generate(
    model,
    processor,
    formatted_prompt,
    image=["./table.png"],
    max_tokens=2048,
    verbose=True,
)

GLM-OCR Result

<table class="table table-bordered"><thead><tr><th></th><th>GLM-OCR</th><th>PaddleOCR-VL-1.5</th><th>Deepseek-OCR2</th><th>MinerU2.5</th><th>dots.ocr</th><th>Gemini-3-Pro</th><th>GPT-5.2-2025-12-11</th></tr></thead><tbody><tr><td></td><td>Specialized VLM</td><td>Specialized VLM</td><td>Specialized VLM</td><td>Specialized VLM</td><td>Specialized VLM</td><td>General VLM</td><td>General VLM</td></tr><tr><td>Document Parsing</td><td></td><td></td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>OmniDocBench v1.5</td><td>94.6</td><td>94.5</td><td>91.1</td><td>90.7</td><td>88.4</td><td>90.3</td><td>85.4</td></tr><tr><td>Text Recognition</td><td></td><td></td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>OmniDocBench v1.5</td><td>94.6</td><td>94.5</td><td>91.1</td><td>90.7</td><td>88.4</td><td>90.3</td><td>85.4</td></tr><tr><td>Information Extraction</td><td></td><td></td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>Nanonets-KIE</td><td>93.7</td><td>-</td><td>-</td><td>-</td><td>-</td><td>95.2</td><td>87.5</td></tr><tr><td>Handwritten-Forms</td><td>86.1</td><td>-</td><td>-</td><td>-</td><td>-</td><td>94.5</td><td>78.2</td></tr></tbody></table>

PaddleOCR-VL-1.5 Result:

MinerU VLM Result:

Expected behavior / 期待表现

Can correctly perform table recognition