torchtorchkimtorch commited on
Commit
fdf4b05
·
verified ·
1 Parent(s): 3179ba3

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +42 -0
README.md CHANGED
@@ -19,3 +19,45 @@ configs:
19
  - split: train
20
  path: data/train-*
21
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
19
  - split: train
20
  path: data/train-*
21
  ---
22
+
23
+ # Muk-Jji-Bba Dataset (SquidGame Series)
24
+
25
+ **Note**: Please do not use this dataset for training purposes.
26
+
27
+ ## Overview
28
+
29
+ The "Muk-Jji-Bba" dataset is the first in the SquidGame series, designed to evaluate whether models can understand human behavior. This dataset specifically focuses on the game of Muk-Jji-Bba, a variation of Rock-Paper-Scissors widely played in Korea.
30
+
31
+ ### How Muk-Jji-Bba Works:
32
+ - The attacker tries to match their gesture with the defender’s to win the game.
33
+ - If both players show the same gesture in the next round, the attacker wins.
34
+ - If the attacker’s next gesture loses to the defender’s, the roles switch, and the defender becomes the new attacker.
35
+ - If the attacker’s next gesture beats the defender’s, the attacker keeps their role and the game continues.
36
+
37
+ The dataset includes 4 rounds per situation. If there is no winner in the final round, the result is a "Tie." The model must choose between Player A (1), Player B (2), or Tie (3).
38
+
39
+ The labels are evenly distributed across the three possible outcomes.
40
+
41
+ ## Model Performance
42
+
43
+ | Model | Accuracy | F1 Score |
44
+ |---------------------|----------|----------|
45
+ | Llama 3.1 (instruct) | 0.8083 | 0.8121 |
46
+ | Llama 3 | 0.775 | 0.7781 |
47
+ | Solar | 0.7541 | 0.7592 |
48
+ | Qwen | 0.7833 | 0.7877 |
49
+ | **Yi-9b-chat (top)** | 0.8395 | 0.8426 |
50
+
51
+ ## About the Labels
52
+ The labels in the dataset represent the correct outcome for each round:
53
+ - Player A wins (1)
54
+ - Player B wins (2)
55
+ - Tie (3)
56
+
57
+ The labels are evenly distributed among the three outcomes to ensure balance.
58
+
59
+ ## Stay Tuned
60
+ Look forward to the next series in the SquidGame dataset!
61
+
62
+ (Evaluation code will be updated soon.)
63
+