Skip to main content

Overview

The VLM Run Feedback API enables you to collect and submit feedback on model predictions to continuously improve accuracy and performance. This feedback data is essential for fine-tuning vlm-1 models to better serve your specific use cases and domains.

Why Feedback Matters

Providing feedback on model predictions serves several critical purposes:
  • Model Fine-tuning: Feedback data is used to fine-tune models for improved accuracy on your specific data patterns
  • Performance Optimization: Helps identify areas where the model needs improvement
  • Domain Adaptation: Enables the model to better understand your industry-specific requirements
  • Quality Assurance: Provides a mechanism to flag incorrect or problematic predictions
1

Submit Feedback

After receiving a prediction, you can submit feedback to help improve future model performance.
from vlmrun.client import VLMRun

client = VLMRun(api_key="<your-api-key>")

feedback_response = client.feedback.submit(
    request_id="<request-id>",
    response={
      "name": "John Doe",
      "date_of_birth": "1955-01-01",
      "email": "john@doe.com"
    },
    notes="The extraction was accurate and captured all key information"
)
print(f"Feedback submitted: {feedback_response.id}")
2

Retrieve Feedback

You can retrieve all feedback associated with a specific prediction.
feedback_list = client.feedback.get("<request-id>")
for feedback in feedback_list:
    print(f"Feedback ID: {feedback.id}")
    print(f"Response: {feedback.response}")
    print(f"Notes: {feedback.notes}")

Fine-tuning with Feedback

The feedback you provide is used to create fine-tuned models that perform better on your specific use cases. This process involves:
  1. Data Collection: Feedback is aggregated across your organization
  2. Model Training: Fine-tuned models are created using your feedback data
  3. Performance Improvement: Updated models show improved accuracy on similar tasks

Best Practices

  • Be Specific: Provide detailed feedback about what was correct or incorrect
  • Use Structured Data: Include ratings, categories, and specific metrics when possible
  • Add Context: Use the notes field to explain your reasoning
  • Consistent Feedback: Maintain consistent criteria across your team for better model training
Fine-tuning capabilities are currently only available for our enterprise-tier customers. If you are interested in using this feature, please contact us.