All submissions will undergo an initial screening to verify the correct and consistent usage of the NdLinear module within the model architecture. Only submissions that pass this correctness check will proceed to the next evaluation stage, where they will be assessed using a comprehensive set of metrics designed to capture both model effectiveness and efficiency.

- **The key evaluation metric is the Top-1 accuracy your model achieves.**

- **Model Shrinkage (%):**
 Model Shrink (%) = ((Your Model Parameter Count - Baseline Model Parameter Count) / Baseline Model Parameter Count) * 100

  This measures the reduction in model size, as indicated by the decrease in the number of parameters, compared to the baseline reference model.

- **Model Complexity:**  
  In addition to the parameter count, we will also consider the computational complexity of the model, typically measured in FLOPs (floating-point operations). Lower complexity with high performance is strongly encouraged.

- **Creativity and Novelty:**  
  The extent to which your approach introduces original ideas, novel adaptations, or creative solutions in leveraging NdLinear will be qualitatively assessed.

**Submission Requirements:**

- Each submission must include detailed instructions along with fully functional scripts for model training, testing, and evaluation.  
- Ensure that your documentation is clear enough for others to run your code and replicate your results without ambiguity.

Submissions that fail to meet the correctness criteria in using the NdLinear module will not advance for further evaluation. We encourage thorough documentation and well-commented code to facilitate the review process and maximize reproducibility.