🚀 LLM Router

User

Test Prompt Routing

0 / 4096

Confidence

Latency

Reason

Quick Stats

Requests Today 0

Small LLM Routes 0

Large LLM Routes 0

Avg Latency 0ms

Live System Logs

Label Distribution

Text Length Distribution

Complexity Indicators

Dataset Statistics

Total Samples 15,306

Small LLM (0) 6,370

Large LLM (1) 8,936

Avg Text Length 78 chars

Model Performance

Training History

Confusion Matrix

6,250

True Negative

120

False Positive

189

False Negative

8,747

True Positive

Key Metrics

Accuracy 97.8%

F1 Score 0.96

Precision 0.97

Recall 0.96

Model Architecture

🧠

DistilBERT

Base Model

🎯

Binary Classification

Task Type

⚡

66M

Parameters

Training Configuration

                                epochs
                                2
                            
                                batch_size
                                8
                            
                                max_length
                                128
                            
                                learning_rate
                                5e-5
                            
                                optimizer
                                AdamW

Why DistilBERT?

40% smaller than BERT
60% faster inference
97% of BERT's performance
Perfect for real-time routing
Low latency (<50ms)

Cross-Validation Results

Robustness Testing

Simple Queries 98.2%

Complex Queries 97.1%

Code Prompts 96.5%

Edge Cases 94.3%

Generalization Analysis

✓

Unseen Patterns

✓

Different Domains

✓

Varying Lengths

✓

Multi-language