Skip to main content
POST
/
training
/
jobs
{
  "job_id": "<string>",
  "judge_id": "<string>",
  "training_type": "<string>",
  "status": "<string>",
  "feedback_data_summary": {
    "total_feedback_count": 123,
    "correction_count": 123,
    "rating_count": 123,
    "date_range": {
      "earliest": {},
      "latest": {}
    }
  },
  "estimated_duration": 123,
  "created_at": {},
  "created_by": "<string>"
}

​
Overview

Creates a new training job to improve judge performance based on accumulated feedback. Training jobs process user corrections and feedback to generate improved judge models with better accuracy and consistency.

​
Request Body

​
judge_id
string
required
ID of the judge to train
​
training_type
string
required
Type of training to perform
  • incremental - Incremental training with recent feedback
  • full_retrain - Complete retraining from all feedback
  • fine_tune - Fine-tuning specific criteria
  • validation - Validation-only run (no model updates)
​
training_config
object
required
Training configuration parameters
​
execution_options
object
Job execution options

​
Response

​
job_id
string
required
Unique identifier for the training job
​
judge_id
string
required
ID of the judge being trained
​
training_type
string
required
Type of training being performed
​
status
string
required
Initial job status
  • created - Job created but not started
  • queued - Job queued for execution
  • preparing - Preparing training data
  • training - Training in progress
  • validating - Validating results
  • completed - Training completed successfully
  • failed - Training failed
  • cancelled - Job was cancelled
​
feedback_data_summary
object
required
Summary of feedback data available for training
​
estimated_duration
integer
Estimated training duration in minutes
​
created_at
datetime
required
Job creation timestamp
​
created_by
string
required
User ID who created the job

​
Examples

curl -X POST https://api.judge.example.com/training/jobs \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "judge_id": "jdg_xyz789",
    "training_type": "incremental",
    "training_config": {
      "feedback_cutoff_date": "2024-01-01T00:00:00Z",
      "minimum_feedback_count": 20,
      "validation_split": 0.2,
      "learning_rate": 0.001,
      "max_epochs": 50,
      "early_stopping": true
    },
    "execution_options": {
      "priority": "high",
      "auto_deploy": false,
      "backup_current_version": true,
      "notification_email": "[email protected]",
      "webhook_url": "https://your-app.com/webhooks/training"
    }
  }'

curl -X POST https://api.judge.example.com/training/jobs \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "judge_id": "jdg_xyz789",
    "training_type": "fine_tune",
    "training_config": {
      "target_criteria": ["crit_clarity", "crit_accuracy"],
      "minimum_feedback_count": 15,
      "learning_rate": 0.0005,
      "max_epochs": 30
    },
    "execution_options": {
      "priority": "normal",
      "auto_deploy": true
    }
  }'

curl -X POST https://api.judge.example.com/training/jobs \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "judge_id": "jdg_new001",
    "training_type": "incremental",
    "training_config": {
      "minimum_feedback_count": 50
    }
  }'