🔍 ERNIE 4.5 Model Comparison Tool

Compare how different ERNIE models process your text with detailed token-level analysis and visualizations.
1 10

Token-Level Analysis

Try these examples:
Input Text Top-k Predictions

How to Interpret Results

  1. Model Summary Chart: Shows which model has higher overall confidence for your input text
  2. Token-Level Confidence Chart: Compares how confident each model is for each token in your text
  3. Log Probability Trend Chart: Shows how log probability changes across tokens (closer to 0 is better)
  4. Token-Level Analysis Table: Detailed breakdown of predictions for each token

Model Differences:

  • ERNIE-4.5-PT: Instruction-tuned model, better at following complex instructions
  • ERNIE-4.5-Base-PT: Base model, better at general language patterns

Note: The first token of your input doesn't have a prediction because there's no preceding context to predict from.