Perfect dataset: Clean unicode artifacts + Add 35 RAG Context-Grounding samples

Add Epistemic Humility & RAG Context-Grounding alignment samples

Apply Option 1: Diversify openings + Structure multi-branch long fatwas

Deep clean web URLs, nan tokens, and ellipses artifacts

Perfect MAFQA Dataset: Complete unabridged premises + Step 4 synthesis + Typo fixes

Publish Perfected MAFQA Multi-Hop Dataset Card

Retain ONLY Pure MAFQA Multi-Hop Dataset (388 samples)

Publish Pure MAFQA Multi-Hop Dataset Card (388 verified samples)

Sanitize any remaining referral regex leakage

Integrate MAFQA Multi-Hop Dataset (388 samples) with QCRI Purified Dataset (4,042 total)

Publish Faqih Multi-Hop & Fara'id SFT Dataset Card (4,042 verified samples)

Overhaul dataset: True Multi-Step CoT + Domain Personas + Context Filtering (3,656 samples)

Overhaul dataset: True Multi-Step CoT + Domain Personas + Context Filtering (3,656 samples)

Publish Faqih True Multi-Step Reasoning SFT Dataset Card

Purge uncurated samples: Retain ONLY 100% verified QCRI Academic Dataset (3,670 samples)

Update Dataset Card to Pure QCRI Verified Academic Dataset (3.67k)

Upload faqih_gold_sft_train.jsonl (11.1k Master Gold Dataset)

Upload faqih_gold_sft_val.jsonl (11.1k Master Gold Dataset)

Upload faqih_gold_sft_train.jsonl (11.1k Master Gold Dataset)

Add comprehensive Dataset Card for 11.1k Master Gold Dataset

initial commit