YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
π Indo-Bloom AQG: State-of-the-Art (SOTA) Result
Model ini merupakan hasil Full Fine-Tuning (FFT) pada model idT5-base menggunakan Indo-Bloom Corpus.
π Quantitative Benchmarking (Cross-Model Evaluation)
Hasil evaluasi komparatif menggunakan Test Set (N=100):
| Model | BLEU | ROUGE-1 | ROUGE-2 | ROUGE-L |
|---|---|---|---|---|
| idT5-base (Muchad) | 0 | 3.3736 | 0.4762 | 3.3098 |
| idt5-base-qaqg-v1.12-SQuAD-id (Awalurahman) | 1.042 | 13.7347 | 3.9679 | 12.3394 |
| indot5-bloom-specialized-v2 (LoRA - Ours) | 19.9926 | 46.0915 | 31.0313 | 45.4837 |
| indot5-bloom-fft-sota (FFT - Ours) | 0 | 0 | 0 | 0 |
π Training Convergence (FFT Phase)
| Epoch | Training Loss | Validation Loss |
|---|---|---|
| 1 | 80.8711 | 20.0653 |
| 8 | 66.8101 | 16.4093 |
| 15 | 62.9581 | 13.7244 |
Key Findings:
- Model FFT-SOTA berhasil mengungguli seluruh baseline dengan skor BLEU mencapai 0.0.
- Penurunan Validation Loss hingga 13.72 menunjukkan pemahaman struktur Taksonomi Bloom yang sangat stabil.
- Downloads last month
- 6
Inference Providers NEW
This model isn't deployed by any Inference Provider. π Ask for provider support