add evaluation/ directory

Author: Fabio Salerno

evaluation/README.md

@@ -0,0 +1,41 @@
+# Evaluation Directory Documentation
+
+This directory contains code, data, and results for data extraction experiments. The evaluation is organized into three main areas: forgetting attacks, memorization attacks, and data inspection.
+
+## Directory Structure
+
+```
+evaluation/
+├── forgetting/
+│   ├── experiments/
+│   └── results/
+├── memorization/
+│   ├── experiments/
+│   └── results/
+└── data-inspection/
+    └── data/
+```
+
+## Forgetting Analysis
+Pre-training code attack experiments on base and fine-tuned models (RQ1-2).
+
+- **experiments/**: Contains data extraction experiment code and extracted code datasets
+- **results/**: Contains generated tables and plots of experimental results
+
+## Memorization Analysis
+Fine-tuning code attack experiments on fine-tuned models (RQ3).
+
+- **experiments/**: Contains data extraction experiment code and extracted code datasets  
+- **results/**: Contains generated tables and plots of experimental results
+
+## Data Inspection
+Code data inspection for memorized code samples (RQ4).
+
+### Files
+- **data/**: Contains tagged dataset files
+- **manual_tag.ipynb**: Jupyter notebook for:
+  - Dataset preparation for manual labeling
+  - Inter-rater agreement statistics
+- **rq4.ipynb**: Jupyter notebook containing tables and analysis for Research Question 4
+
+The evaluation directory provides a comprehensive framework for analyzing model behavior through data extraction attacks and detailed code inspection.