File size: 1,722 Bytes
cb592f0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
---
license: apache-2.0
tags:
- question-answering
- complexity-classification
- distilbert
datasets:
- wesley7137/question_complexity_classification
---

# question-complexity-classifier

馃 Fine-tuned DistilBERT model for classifying question complexity (Simple vs Complex)

## Model Details

### Model Description
- **Architecture:** DistilBERT base uncased
- **Fine-tuned on:** Question Complexity Classification Dataset
- **Language:** English
- **License:** Apache 2.0
- **Max Sequence Length:** 128 tokens

## Uses

```python
from transformers import pipeline

classifier = pipeline(
    "text-classification",
    model="grahamaco/question-complexity-classifier",
    tokenizer="grahamaco/question-complexity-classifier",
    truncation=True,
    max_length=128  # Matches training config
)

result = classifier("Explain quantum computing in simple terms")
# Output example: {'label': 'COMPLEX', 'score': 0.97}
```

## Training Details

- **Epochs:** 5
- **Batch Size:** 32 (global)
- **Learning Rate:** 2e-5
- **Train/Val/Test Split:** 80/10/10 (stratified)
- **Early Stopping:** Patience of 2 epochs

## Evaluation Results

| Metric | Value |
|--------|-------|
| Accuracy | 0.92 |
| F1 Score | 0.91 |

## Performance

| Metric | Value |
|--------|-------|
| Inference Latency | 15.2ms (CPU) |
| Throughput | 68.4 samples/sec (GPU) |

## Ethical Considerations
This model is intended for educational content classification only. Developers should:
- Regularly audit performance across different question types
- Monitor for unintended bias in complexity assessments
- Provide human-review mechanisms for high-stakes classifications
- Validate classifications against original context when used with RAG systems