Generated Question Outputs and DeepEval Evaluation Results
收藏官方服务:
资源简介:
This dataset contains generated question outputs and DeepEval evaluation results from a systematic ablation study across eight pipeline configurations (C0–C7), along with extended baseline comparisons with Qwen2.5-7B-INT4 and GPT-4o-mini. Base model: Qwen2.5-3B-Instruct (no fine-tuning)Evaluation framework: DeepEval (GPT-4o-mini as judge)Total questions: 400 (50 × 8 configurations) + 100 (extended baseline) The following documents were used as RAG knowledge base: Buku Matematika Kelas 4 SD — Kemendikbudristek, 2022 Lembar Kerja Siswa (LKS) Matematika Kelas 4 SD Bank Soal Matematika Kelas 4 SD
提供机构:
Zenodo创建时间:
2026-07-12



