mbayan commited on
Commit
b77ffea
·
verified ·
1 Parent(s): bdff8da

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +18 -15
README.md CHANGED
@@ -60,7 +60,7 @@ The dataset is provided in a structured format:
60
  ```python
61
  from datasets import load_dataset
62
 
63
- dataset = load_dataset("your-username/your-dataset-name")
64
  print(dataset)
65
  ```
66
 
@@ -76,18 +76,21 @@ Each sample includes:
76
 
77
  We evaluated the dataset using **LLaMA-based models** with different configurations. Below is a summary of our findings:
78
 
79
- | Metric | LLaMA-3.2-3B | LLaMA-3.1-8B | LLaMA-3.2-3B-1S | LLaMA-3.2-3B-FT |
80
- |---------------|-------------|-------------|----------------|----------------|
81
- | Coherence | 2.69 | 5.49 | 4.52 | 5.31 |
82
- | Brevity | 1.99 | 4.30 | 3.76 | 4.98 |
83
- | Legal Language | 3.66 | 6.69 | 5.18 | 5.87 |
84
- | Faithfulness | 3.00 | 5.99 | 4.00 | 5.24 |
85
- | Clarity | 2.90 | 5.79 | 4.99 | 5.94 |
86
- | Consistency | 3.04 | 5.93 | 5.14 | 6.16 |
87
- | **Avg. Score** | 3.01 | 5.89 | 4.66 | 5.68 |
88
- | ROUGE-1 | 0.08 | 0.12 | 0.29 | 0.34 |
89
- | ROUGE-2 | 0.02 | 0.04 | 0.19 | 0.25 |
90
- | BLEU | 0.01 | 0.02 | 0.11 | 0.14 |
 
 
 
91
 
92
  ### **Key Findings**
93
  - Fine-tuned smaller models (**LLaMA-3.2-3B-FT**) achieve performance **comparable to larger models** (LLaMA-3.1-8B).
@@ -101,7 +104,7 @@ To use the dataset in your research, load it as follows:
101
  ```python
102
  from datasets import load_dataset
103
 
104
- dataset = load_dataset("your-username/your-dataset-name")
105
 
106
  # Access train and test splits
107
  train_data = dataset["train"]
@@ -112,7 +115,7 @@ test_data = dataset["test"]
112
 
113
  The full implementation, including preprocessing scripts and model training code, is available in our GitHub repository:
114
 
115
- 🔗 **[GitHub: MohamedBayan/to_be_released](https://github.com/MohamedBayan/to_be_released)**
116
 
117
  ## Citation
118
 
 
60
  ```python
61
  from datasets import load_dataset
62
 
63
+ dataset = load_dataset("mbayan/Arabic-LJP")
64
  print(dataset)
65
  ```
66
 
 
76
 
77
  We evaluated the dataset using **LLaMA-based models** with different configurations. Below is a summary of our findings:
78
 
79
+ | **Metric** | **LLaMA-3.2-3B** | **LLaMA-3.1-8B** | **LLaMA-3.2-3B-1S** | **LLaMA-3.2-3B-FT** | **LLaMA-3.1-8B-FT** |
80
+ |--------------------------|------------------|------------------|---------------------|---------------------|---------------------|
81
+ | **Coherence** | 2.69 | 5.49 | 4.52 | *6.60* | **6.94** |
82
+ | **Brevity** | 1.99 | 4.30 | 3.76 | *5.87* | **6.27** |
83
+ | **Legal Language** | 3.66 | 6.69 | 5.18 | *7.48* | **7.73** |
84
+ | **Faithfulness** | 3.00 | 5.99 | 4.00 | *6.08* | **6.42** |
85
+ | **Clarity** | 2.90 | 5.79 | 4.99 | *7.90* | **8.17** |
86
+ | **Consistency** | 3.04 | 5.93 | 5.14 | *8.47* | **8.65** |
87
+ | **Avg. Qualitative Score**| 3.01 | 5.89 | 4.66 | *7.13* | **7.44** |
88
+ | **ROUGE-1** | 0.08 | 0.12 | 0.29 | *0.50* | **0.53** |
89
+ | **ROUGE-2** | 0.02 | 0.04 | 0.19 | *0.39* | **0.41** |
90
+ | **BLEU** | 0.01 | 0.02 | 0.11 | *0.24* | **0.26** |
91
+ | **BERT** | 0.54 | 0.58 | 0.64 | *0.74* | **0.76** |
92
+
93
+ **Caption**: A comparative analysis of performance across different LLaMA models. The model names have been abbreviated for simplicity: **LLaMA-3.2-3B-Instruct** is represented as LLaMA-3.2-3B, **LLaMA-3.1-8B-Instruct** as LLaMA-3.1-8B, **LLaMA-3.2-3B-Instruct-1-Shot** as LLaMA-3.2-3B-1S, **LLaMA-3.2-3B-Instruct-Finetuned** as LLaMA-3.2-3B-FT, and **LLaMA-3.1-8B-Finetuned** as LLaMA-3.1-8B-FT.
94
 
95
  ### **Key Findings**
96
  - Fine-tuned smaller models (**LLaMA-3.2-3B-FT**) achieve performance **comparable to larger models** (LLaMA-3.1-8B).
 
104
  ```python
105
  from datasets import load_dataset
106
 
107
+ dataset = load_dataset("mbayan/Arabic-LJP")
108
 
109
  # Access train and test splits
110
  train_data = dataset["train"]
 
115
 
116
  The full implementation, including preprocessing scripts and model training code, is available in our GitHub repository:
117
 
118
+ 🔗 **[GitHub: MohamedBayan/to_be_released](https://github.com/MohamedBayan/Arabic-Legal-Judgment-Prediction)**
119
 
120
  ## Citation
121