kth8/Qwen3.5-4B-MMLU-Pro-benchmark

--- license: apache-2.0 language: - en base_model: Qwen/Qwen3.5-4B datasets: - TIGER-Lab/MMLU-Pro --- Benchmark of [Qwen/Qwen3.5-4B](https://huggingface.co/Qwen/Qwen3.5-4B) against [TIGER-Lab/MMLU-Pro](https://huggingface.co/datasets/TIGER-Lab/MMLU-Pro) dataset. Accuracy: 74.5% with Python tool. | Metric | Value | |----------------------|---------------| | **Correct** | 745 | | **Incorrect** | 254 | | **Errors** | 1 | | **Total samples** | 1000 | | **Python tool calls**| 1182 | | **Total completion tokens** | 2,318,269 | Raw stats: ```json { "accuracy": 0.745, "correct": 745, "incorrect": 254, "error": 1, "total": 1000, "python_tool_calls": 1182, "completion_tokens": 2318269 } ```