GLM

• zai-org/GLM-4.1V-Thinking
  • 9B, 64K context, Vision, Reasoning
  • 基于 GLM-4-9B-0414
  • glmv_reward
• <|begin_of_box|>, <|end_of_box|>
  • 核心答案
  • 可能会导致格式错误
  • https://github.com/zai-org/GLM-4.1V-Thinking/issues/79
• <think>
• </think>
• <answer>
• </answer>

glm-ocr

default_prompt: >
  Recognize the text in the image and output in Markdown format.
  Preserve the original layout (headings/paragraphs/tables/formulas).
  Do not fabricate content that does not exist in the image.
task_prompt_mapping:
  text: 'Text Recognition:'
  table: 'Table Recognition:'
  formula: 'Formula Recognition:'

信息提取

请按下列JSON格式输出图中信息:
{
    "id_number": "",
    "last_name": "",
    "first_name": "",
    "date_of_birth": "",
    "address": {
        "street": "",
        "city": "",
        "state": "",
        "zip_code": ""
    },
    "dates": {
        "issue_date": "",
        "expiration_date": ""
    },
    "sex": ""
}