Individualized Education Program (IEP) Generation on Formal n=10 (test)
100Schema ComplianceGPT-5.4
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GPT-5.4Vendor=OpenAI, mode=zero-shot2026.06 | 100 | 100 | 72.6 | |
| DeepSeek-V3.2Vendor=DeepSeek, mode=zero-shot2026.06 | 100 | 100 | 70.3 | |
| Breeze-7BVendor=MediaTek, mode=zero-shot (B1)2026.06 | 100 | — | 65.3 | |
| Breeze-7B + QLoRAVendor=MediaTek, GCD=ON, ablation=B22026.06 | 100 | — | 75.2 | |
| Breeze-7B + QLoRAVendor=MediaTek, GCD=OFF (no-GCD), mode=free-form2026.06 | 100 | — | 77.9 | |
| Gemini-3-Flash-PreviewVendor=Google, mode=zero-shot2026.06 | 90 | 90 | 70.3 | |
| Llama-4-MaverickVendor=Meta, mode=zero-shot2026.06 | 90 | 90 | 70 |