Upload README.md with huggingface_hub

README.md

@@ -1,84 +1,87 @@
----
-license: mit
-task_categories:
-- question-answering
-- text-generation
-language:
-- en
-tags:
-- multi-hop
-- tool-calling
-- reasoning
-- qa
-size_categories:
-- 1K<n<10K
----
-
 # Multi-Hop Tool Execution Dataset
 
-This dataset contains 1,100 multi-hop questions with step-by-step tool execution and answers.
+This dataset contains multi-hop questions with tool execution results, designed for training and evaluating AI systems on complex reasoning tasks.
 
 ## Dataset Description
 
-- **Total Questions**: 1,100
-- **Question Complexity**: 5-15 steps
-- **Questions per Length**: 100 each
-- **Tools Used**: duckduckgo_search, fetch_url_content, python
-- **Splits**: Train (990), Validation (110)
-
-## Dataset Structure
-
-Each example contains:
-- `id`: Unique question identifier
-- `main_question`: The overarching multi-hop question
-- `sub_questions`: JSON string of sub-questions for each step
-- `domain`: Question domain (e.g., financial_analysis, technology_metrics)
-- `tool_chain`: JSON string of tools used in sequence
-- `question_length`: Number of steps in the question
-- `batch_number`: Batch number this question belongs to
-- `batch_length`: Length category of this question (5-15)
-- `question_in_batch`: Position of this question within its batch
-- `step_by_step_execution`: JSON string of detailed execution results
-- `final_answer`: The computed final answer
-
-## Usage
+This dataset contains **1100 questions** ranging from 5-15 steps, each requiring multiple tool interactions to reach the final answer.
+
+### Tools Used
+- **duckduckgo_search**: Web search functionality
+- **fetch_url_content**: URL content retrieval  
+- **python**: Mathematical calculations and data processing
+
+### Dataset Schema
+
+| Field | Type | Description |
+|-------|------|-------------|
+| `main_question` | string | The overarching question requiring multi-hop reasoning |
+| `sub_questions` | list | List of sub-questions for each step |
+| `domain` | string | Domain category (e.g., healthcare_metrics, technology_metrics) |
+| `tool_chain` | list | Sequence of tools used |
+| `question_length` | int64 | Number of steps in the question |
+| `batch_number` | int64 | Batch number for organization |
+| `batch_length` | int64 | Length category of the batch |
+| `question_in_batch` | int64 | Position within the batch |
+| `step_by_step_execution` | list | Detailed execution results for each step |
+| `final_answer` | string | The computed final answer |
+
+### Statistics
+
+- **Total Questions**: 1100
+- **Question Lengths**: 5-15 steps
+- **Domains**: 9 different domains
+- **Tool Usage Distribution**:
+  - Search operations: ~50%
+  - Content fetching: ~30% 
+  - Python calculations: ~20%
+
+### Usage
 
 ```python
 from datasets import load_dataset
-import json
 
 # Load the dataset
 dataset = load_dataset("Anna4242/multihop-tool-execution-dataset")
 
-# Access train split
+# Access train/validation splits
 train_data = dataset["train"]
-print(f"Train examples: {len(train_data)}")
+val_data = dataset["validation"]
 
 # Example usage
-example = train_data[0]
-print(f"Question: {example['main_question']}")
-print(f"Steps: {example['question_length']}")
+for example in train_data:
+    question = example["main_question"]
+    tools = example["tool_chain"]
+    answer = example["final_answer"]
+    print(f"Question: {question}")
+    print(f"Tools: {' → '.join(tools)}")
+    print(f"Answer: {answer}")
+```
 
-# Parse JSON fields
-sub_questions = json.loads(example['sub_questions'])
-tool_chain = json.loads(example['tool_chain'])
-execution_details = json.loads(example['step_by_step_execution'])
+### Data Quality
 
-print(f"Sub-questions: {len(sub_questions)}")
-print(f"Tools: {tool_chain}")
-```
+All questions in this dataset have been executed and validated:
+- Each question includes detailed step-by-step execution results
+- Tool outputs are captured for each step
+- Final answers are computed from the execution chain
+
+### Citation
 
-## Citation
+If you use this dataset, please cite:
 
 ```bibtex
-@dataset{multihop_tool_execution,
+@dataset{multihop_tool_execution_2024,
   title={Multi-Hop Tool Execution Dataset},
   author={Anna4242},
-  year={2025},
+  year={2024},
   url={https://huggingface.co/datasets/Anna4242/multihop-tool-execution-dataset}
 }
 ```
 
-## License
+### License
 
 This dataset is released under the MIT License.
+
+---
+
+Generated on 2025-08-12 06:03:05