Update README.md

# MultiReasoningQA

A unified dataset of 10,000 QA samples across 5 reasoning types:
- **Mathematical** (from GSM8K)
- **Multihop** (from HotpotQA)
- **Reading comprehension** (from DROP)
- **Commonsense** (from CommonsenseQA)
- **Logical** (from ReClor)

## Dataset Structure

**Columns**:
- `question` — The question text (may include context)
- `answer` — Correct answer
- `reasoning_type` — One of: `mathematical`, `multihop`, `reading_comp`, `commonsense`, `logical`
- `source` — Original dataset name
- `id` — Unique ID per row

**Splits**:
- `train`: 7847 rows
- `validation`: 980 rows
- `test`: 980 rows

README.md

@@ -54,4 +54,10 @@ configs:
     path: data/validation-*
   - split: test
     path: data/test-*
----
+task_categories:
+- question-answering
+language:
+- en
+size_categories:
+- 1K<n<10K
+---