Update README.md
Browse files
README.md
CHANGED
@@ -39,27 +39,27 @@ In this phase, the annealed models were fine-tuned on meticulously curated in-ho
|
|
39 |
|
40 |
The models were evaluated on various French language tasks, with results detailed in the table below:
|
41 |
|
42 |
-
| Model
|
43 |
-
|
|
44 |
-
| CroissantLLMBase
|
45 |
-
| SmolLM2-1.7B
|
46 |
-
| Mistral-7B-v0.3
|
47 |
-
| Lucie-7B
|
48 |
-
|
|
49 |
-
| Llama-3.2-1B
|
50 |
-
| Vigogne_Llama-3.2-1B | 0.6338 | 0.2814 | 0.4136 | 0.7647 | 0.5561 | 0.7983 | 0.5747 |
|
51 |
-
|
|
52 |
-
| Qwen2.5-1.5B
|
53 |
-
| Vigogne_Qwen2.5-1.5B | 0.6619 | 0.3122 | 0.4514 | 0.8403 | 0.5393 | 0.8067 | 0.6019 |
|
54 |
-
|
|
55 |
-
| Llama-3.2-3B
|
56 |
-
| Vigogne_Llama-3.2-3B | 0.6760 | 0.3669 | 0.4897 | 0.8403 | 0.6966 | 0.8403 | 0.6496 |
|
57 |
-
|
|
58 |
-
| Qwen2.5-3B
|
59 |
-
| Vigogne_Qwen2.5-3B | 0.6619 | 0.4080 | 0.4922 | 0.8151 | 0.7247 | 0.8235 | 0.6542 |
|
60 |
-
|
|
61 |
-
| Llama-3.1-8B
|
62 |
-
| Vigogne_Llama-3.1-8B | 0.6760 | 0.4148 | 0.5240 | 0.8067 | 0.7977 | 0.8235 | 0.6738 |
|
63 |
|
64 |
### Reproducing Results
|
65 |
|
|
|
39 |
|
40 |
The models were evaluated on various French language tasks, with results detailed in the table below:
|
41 |
|
42 |
+
| Model | Reading Comp | ARC Challenge | HellaSwag | Grammar | BoolQA | French Bench Vocab | Avg |
|
43 |
+
| ------------------------ | ------------ | ------------- | --------- | ------- | ------ | ------------------ | ------ |
|
44 |
+
| CroissantLLMBase | 0.6197 | 0.2258 | 0.3918 | 0.7815 | 0.4887 | 0.7815 | 0.5481 |
|
45 |
+
| SmolLM2-1.7B | 0.5211 | 0.2592 | 0.3327 | 0.6134 | 0.5506 | 0.5966 | 0.4789 |
|
46 |
+
| Mistral-7B-v0.3 | 0.6619 | 0.3806 | 0.4729 | 0.7563 | 0.4943 | 0.7815 | 0.5912 |
|
47 |
+
| Lucie-7B | 0.6338 | 0.4097 | 0.4925 | 0.7983 | 0.5505 | 0.8151 | 0.6166 |
|
48 |
+
| | | | | | | | |
|
49 |
+
| Llama-3.2-1B | 0.5493 | 0.2387 | 0.3548 | 0.6891 | 0.5674 | 0.7563 | 0.5259 |
|
50 |
+
| **Vigogne_Llama-3.2-1B** | 0.6338 | 0.2814 | 0.4136 | 0.7647 | 0.5561 | 0.7983 | 0.5747 |
|
51 |
+
| | | | | | | | |
|
52 |
+
| Qwen2.5-1.5B | 0.5915 | 0.3045 | 0.3821 | 0.7563 | 0.7191 | 0.7479 | 0.5836 |
|
53 |
+
| **Vigogne_Qwen2.5-1.5B** | 0.6619 | 0.3122 | 0.4514 | 0.8403 | 0.5393 | 0.8067 | 0.6019 |
|
54 |
+
| | | | | | | | |
|
55 |
+
| Llama-3.2-3B | 0.6760 | 0.3550 | 0.4315 | 0.7731 | 0.5000 | 0.7899 | 0.5876 |
|
56 |
+
| **Vigogne_Llama-3.2-3B** | 0.6760 | 0.3669 | 0.4897 | 0.8403 | 0.6966 | 0.8403 | 0.6496 |
|
57 |
+
| | | | | | | | |
|
58 |
+
| Qwen2.5-3B | 0.5774 | 0.3567 | 0.4344 | 0.7563 | 0.8932 | 0.7983 | 0.6361 |
|
59 |
+
| **Vigogne_Qwen2.5-3B** | 0.6619 | 0.4080 | 0.4922 | 0.8151 | 0.7247 | 0.8235 | 0.6542 |
|
60 |
+
| | | | | | | | |
|
61 |
+
| Llama-3.1-8B | 0.7042 | 0.4174 | 0.4881 | 0.7815 | 0.4943 | 0.8067 | 0.6154 |
|
62 |
+
| **Vigogne_Llama-3.1-8B** | 0.6760 | 0.4148 | 0.5240 | 0.8067 | 0.7977 | 0.8235 | 0.6738 |
|
63 |
|
64 |
### Reproducing Results
|
65 |
|