## Evaluation for Part 1

Your submission will be evaluated based on how closely your denoised weights match the ground truth clean weights using Normalized Root Mean Squared Error (NRMSE).
### Normalized Mean Squared Error (NMSE)

**Formula:**

$$
\mathrm{NMSE} = \frac{\sum (y - \hat{y})^2}{\sum y^2}
$$


**Where:**
- y = true values  
- ŷ = predicted values



## Evaluation for Part 2
- You will be evaluated on your ability to correctly translate team-specific encoded phrases
- Evaluation will be partially manual to ensure fairness
- Translation accuracy will be measured on unseen encoded text
    - For this, we will be using:
      - **BLEU Score**: Measures n-gram overlap with reference translations
      - **Levenshtein Distance**: Character-level edit distance
      - **ROUGE Score**: Measures recall of n-grams