Upload README.md with huggingface_hub

README.md

@@ -1,756 +1,717 @@
----
-license: cc-by-nc-4.0
-language:
-- ro
-base_model:
-- meta-llama/Meta-Llama-3-8B-Instruct
-datasets:
-- OpenLLM-Ro/ro_sft_alpaca
-- OpenLLM-Ro/ro_sft_alpaca_gpt4
-- OpenLLM-Ro/ro_sft_dolly
-- OpenLLM-Ro/ro_sft_selfinstruct_gpt4
-- OpenLLM-Ro/ro_sft_norobots
-- OpenLLM-Ro/ro_sft_orca
-- OpenLLM-Ro/ro_sft_camel
-- OpenLLM-Ro/ro_sft_oasst
-- OpenLLM-Ro/ro_sft_ultrachat
-model-index:
-    - name: OpenLLM-Ro/RoLlama3-8b-Instruct-2024-10-09
-      results:
-        - task:
-            type: text-generation
-          dataset:
-            name: RoMT-Bench
-            type: RoMT-Bench
-          metrics:
-            - name: Score
-              type: Score
-              value: 5.38
-        - task:
-            type: text-generation
-          dataset:
-            name: RoCulturaBench
-            type: RoCulturaBench
-          metrics:
-            - name: Score
-              type: Score
-              value: 3.81
-        - task:
-            type: text-generation
-          dataset:
-            name: Romanian_Academic_Benchmarks
-            type: Romanian_Academic_Benchmarks
-          metrics:
-            - name: Average accuracy
-              type: accuracy
-              value: 52.21
-        - task:
-            type: text-generation
-          dataset:
-            name: OpenLLM-Ro/ro_arc_challenge
-            type: OpenLLM-Ro/ro_arc_challenge
-          metrics:
-            - name: Average accuracy
-              type: accuracy
-              value: 47.94
-        - task:
-            type: text-generation
-          dataset:
-            name: OpenLLM-Ro/ro_mmlu
-            type: OpenLLM-Ro/ro_mmlu
-          metrics:
-            - name: Average accuracy
-              type: accuracy
-              value: 53.50
-        - task:
-            type: text-generation
-          dataset:
-            name: OpenLLM-Ro/ro_winogrande
-            type: OpenLLM-Ro/ro_winogrande
-          metrics:
-            - name: Average accuracy
-              type: accuracy
-              value: 66.06
-        - task:
-            type: text-generation
-          dataset:
-            name: OpenLLM-Ro/ro_hellaswag
-            type: OpenLLM-Ro/ro_hellaswag
-          metrics:
-            - name: Average accuracy
-              type: accuracy
-              value: 59.72
-        - task:
-            type: text-generation
-          dataset:
-            name: OpenLLM-Ro/ro_gsm8k
-            type: OpenLLM-Ro/ro_gsm8k
-          metrics:
-            - name: Average accuracy
-              type: accuracy
-              value: 40.16
-        - task:
-            type: text-generation
-          dataset:
-            name: OpenLLM-Ro/ro_truthfulqa
-            type: OpenLLM-Ro/ro_truthfulqa
-          metrics:
-            - name: Average accuracy
-              type: accuracy
-              value: 45.90
-        - task:
-            type: text-generation
-          dataset:
-            name: LaRoSeDa_binary
-            type: LaRoSeDa_binary
-          metrics:
-            - name: Average macro-f1
-              type: macro-f1
-              value: 95.58
-        - task:
-            type: text-generation
-          dataset:
-            name: LaRoSeDa_multiclass
-            type: LaRoSeDa_multiclass
-          metrics:
-            - name: Average macro-f1
-              type: macro-f1
-              value: 61.20
-        - task:
-            type: text-generation
-          dataset:
-            name: LaRoSeDa_binary_finetuned
-            type: LaRoSeDa_binary_finetuned
-          metrics:
-            - name: Average macro-f1
-              type: macro-f1
-              value: 96.46
-        - task:
-            type: text-generation
-          dataset:
-            name: LaRoSeDa_multiclass_finetuned
-            type: LaRoSeDa_multiclass_finetuned
-          metrics:
-            - name: Average macro-f1
-              type: macro-f1
-              value: 87.26
-        - task:
-            type: text-generation
-          dataset:
-            name: WMT_EN-RO
-            type: WMT_EN-RO
-          metrics:
-            - name: Average bleu
-              type: bleu
-              value: 22.92
-        - task:
-            type: text-generation
-          dataset:
-            name: WMT_RO-EN
-            type: WMT_RO-EN
-          metrics:
-            - name: Average bleu
-              type: bleu
-              value: 24.28
-        - task:
-            type: text-generation
-          dataset:
-            name: WMT_EN-RO_finetuned
-            type: WMT_EN-RO_finetuned
-          metrics:
-            - name: Average bleu
-              type: bleu
-              value: 27.31
-        - task:
-            type: text-generation
-          dataset:
-            name: WMT_RO-EN_finetuned
-            type: WMT_RO-EN_finetuned
-          metrics:
-            - name: Average bleu
-              type: bleu
-              value: 40.52
-        - task:
-            type: text-generation
-          dataset:
-            name: XQuAD
-            type: XQuAD
-          metrics:
-            - name: Average exact_match
-              type: exact_match
-              value: 18.89
-        - task:
-            type: text-generation
-          dataset:
-            name: XQuAD
-            type: XQuAD
-          metrics:
-            - name: Average f1
-              type: f1
-              value: 31.79
-        - task:
-            type: text-generation
-          dataset:
-            name: XQuAD_finetuned
-            type: XQuAD_finetuned
-          metrics:
-            - name: Average exact_match
-              type: exact_match
-              value: 50.84
-        - task:
-            type: text-generation
-          dataset:
-            name: XQuAD_finetuned
-            type: XQuAD_finetuned
-          metrics:
-            - name: Average f1
-              type: f1
-              value: 65.18
-        - task:
-            type: text-generation
-          dataset:
-            name: STS
-            type: STS
-          metrics:
-            - name: Average spearman
-              type: spearman
-              value: 77.60
-        - task:
-            type: text-generation
-          dataset:
-            name: STS
-            type: STS
-          metrics:
-            - name: Average pearson
-              type: pearson
-              value: 76.86
-        - task:
-            type: text-generation
-          dataset:
-            name: STS_finetuned
-            type: STS_finetuned
-          metrics:
-            - name: Average spearman
-              type: spearman
-              value: 86.70
-        - task:
-            type: text-generation
-          dataset:
-            name: STS_finetuned
-            type: STS_finetuned
-          metrics:
-            - name: Average pearson
-              type: pearson
-              value: 87.09
-        - task:
-            type: text-generation
-          dataset:
-            name: RoMT-Bench
-            type: RoMT-Bench
-          metrics:
-            - name: First turn
-              type: Score
-              value: 6.09
-            - name: Second turn
-              type: Score
-              value: 4.67
-        - task:
-            type: text-generation
-          dataset:
-            name: OpenLLM-Ro/ro_arc_challenge
-            type: OpenLLM-Ro/ro_arc_challenge
-          metrics:
-            - name: 0-shot 
-              type: accuracy
-              value: 46.02
-            - name: 1-shot 
-              type: accuracy
-              value: 47.39
-            - name: 3-shot 
-              type: accuracy
-              value: 47.73
-            - name: 5-shot 
-              type: accuracy
-              value: 48.24
-            - name: 10-shot 
-              type: accuracy
-              value: 48.33
-            - name: 25-shot 
-              type: accuracy
-              value: 49.96
-        - task:
-            type: text-generation
-          dataset:
-            name: OpenLLM-Ro/ro_mmlu
-            type: OpenLLM-Ro/ro_mmlu
-          metrics:
-            - name: 0-shot 
-              type: accuracy
-              value: 51.19
-            - name: 1-shot 
-              type: accuracy
-              value: 53.05
-            - name: 3-shot 
-              type: accuracy
-              value: 54.83
-            - name: 5-shot 
-              type: accuracy
-              value: 54.93
-        - task:
-            type: text-generation
-          dataset:
-            name: OpenLLM-Ro/ro_winogrande
-            type: OpenLLM-Ro/ro_winogrande
-          metrics:
-            - name: 0-shot 
-              type: accuracy
-              value: 64.09
-            - name: 1-shot 
-              type: accuracy
-              value: 66.22
-            - name: 3-shot 
-              type: accuracy
-              value: 66.61
-            - name: 5-shot 
-              type: accuracy
-              value: 67.32
-        - task:
-            type: text-generation
-          dataset:
-            name: OpenLLM-Ro/ro_hellaswag
-            type: OpenLLM-Ro/ro_hellaswag
-          metrics:
-            - name: 0-shot 
-              type: accuracy
-              value: 59.34
-            - name: 1-shot 
-              type: accuracy
-              value: 59.52
-            - name: 3-shot 
-              type: accuracy
-              value: 59.61
-            - name: 5-shot 
-              type: accuracy
-              value: 59.95
-            - name: 10-shot 
-              type: accuracy
-              value: 60.19
-        - task:
-            type: text-generation
-          dataset:
-            name: OpenLLM-Ro/ro_gsm8k
-            type: OpenLLM-Ro/ro_gsm8k
-          metrics:
-            - name: 1-shot 
-              type: accuracy
-              value: 31.31
-            - name: 3-shot 
-              type: accuracy
-              value: 42.23
-            - name: 5-shot 
-              type: accuracy
-              value: 46.93
-        - task:
-            type: text-generation
-          dataset:
-            name: LaRoSeDa_binary
-            type: LaRoSeDa_binary
-          metrics:
-            - name: 0-shot 
-              type: macro-f1
-              value: 92.43
-            - name: 1-shot 
-              type: macro-f1
-              value: 96.23
-            - name: 3-shot 
-              type: macro-f1
-              value: 96.66
-            - name: 5-shot 
-              type: macro-f1
-              value: 97.00
-        - task:
-            type: text-generation
-          dataset:
-            name: LaRoSeDa_multiclass
-            type: LaRoSeDa_multiclass
-          metrics:
-            - name: 0-shot 
-              type: macro-f1
-              value: 61.47
-            - name: 1-shot 
-              type: macro-f1
-              value: 63.77
-            - name: 3-shot 
-              type: macro-f1
-              value: 57.12
-            - name: 5-shot 
-              type: macro-f1
-              value: 62.43
-        - task:
-            type: text-generation
-          dataset:
-            name: WMT_EN-RO
-            type: WMT_EN-RO
-          metrics:
-            - name: 0-shot 
-              type: bleu
-              value: 5.25
-            - name: 1-shot 
-              type: bleu
-              value: 28.62
-            - name: 3-shot 
-              type: bleu
-              value: 29.60
-            - name: 5-shot 
-              type: bleu
-              value: 28.21
-        - task:
-            type: text-generation
-          dataset:
-            name: WMT_RO-EN
-            type: WMT_RO-EN
-          metrics:
-            - name: 0-shot 
-              type: bleu
-              value: 1.95
-            - name: 1-shot 
-              type: bleu
-              value: 24.00
-            - name: 3-shot 
-              type: bleu
-              value: 34.87
-            - name: 5-shot 
-              type: bleu
-              value: 36.31
-        - task:
-            type: text-generation
-          dataset:
-            name: XQuAD_EM
-            type: XQuAD_EM
-          metrics:
-            - name: 0-shot 
-              type: exact_match
-              value: 16.97
-            - name: 1-shot 
-              type: exact_match
-              value: 31.01
-            - name: 3-shot 
-              type: exact_match
-              value: 13.95
-            - name: 5-shot 
-              type: exact_match
-              value: 13.61
-        - task:
-            type: text-generation
-          dataset:
-            name: XQuAD_F1
-            type: XQuAD_F1
-          metrics:
-            - name: 0-shot 
-              type: f1
-              value: 31.29
-            - name: 1-shot 
-              type: f1
-              value: 42.77
-            - name: 3-shot 
-              type: f1
-              value: 24.78
-            - name: 5-shot 
-              type: f1
-              value: 28.30
-        - task:
-            type: text-generation
-          dataset:
-            name: STS_Spearman
-            type: STS_Spearman
-          metrics:
-            - name: 1-shot 
-              type: spearman
-              value: 77.73
-            - name: 3-shot 
-              type: spearman
-              value: 76.78
-            - name: 5-shot 
-              type: spearman
-              value: 78.30
-        - task:
-            type: text-generation
-          dataset:
-            name: STS_Pearson
-            type: STS_Pearson
-          metrics:
-            - name: 1-shot 
-              type: pearson
-              value: 77.25
-            - name: 3-shot 
-              type: pearson
-              value: 75.83
-            - name: 5-shot 
-              type: pearson
-              value: 77.49
-
----
-
-# Model Card for Model ID
-
-*Built with Meta Llama 3*
-
-This model points/is identical to [RoLlama3-8b-Instruct-2024-10-09](https://huggingface.co/OpenLLM-Ro/RoLlama3-8b-Instruct-2024-10-09).
-
-<!-- Provide a quick summary of what the model is/does. -->
-
-RoLlama3 is a family of pretrained and fine-tuned generative text models for Romanian. This is the repository for the **instruct 8B model**. Links to other models can be found at the bottom of this page.
-
-
-## Model Details
-
-### Model Description
-
-<!-- Provide a longer summary of what this model is. -->
-OpenLLM-Ro represents the first open-source effort to build a LLM specialized for Romanian. OpenLLM-Ro developed and publicly releases a collection of Romanian LLMs, both in the form of foundational model and instruct and chat variants.
-
-
-- **Developed by:** OpenLLM-Ro
-<!-- - **Funded by [optional]:** [More Information Needed] -->
-<!-- - **Shared by [optional]:** [More Information Needed] -->
-<!-- - **Model type:** [More Information Needed] -->
-- **Language(s):** Romanian
-- **License:** cc-by-nc-4.0
-- **Finetuned from model:** [Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct)
-- **Trained using:** [RoAlpaca](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_alpaca), [RoAlpacaGPT4](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_alpaca_gpt4), [RoDolly](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_dolly), [RoSelfInstruct](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_selfinstruct_gpt4), [RoNoRobots](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_norobots), [RoOrca](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_orca), [RoCamel](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_camel), [RoOpenAssistant](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_oasst), [RoUltraChat](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_ultrachat)
-
-
-### Model Sources
-
-<!-- Provide the basic links for the model. -->
-
-- **Repository:** https://github.com/OpenLLM-Ro/LLaMA-Factory
-- **Paper:** https://arxiv.org/abs/2406.18266
-
-## Intended Use
-
-### Intended Use Cases
-
-RoLlama3 is intented for research use in Romanian. Base models can be adapted for a variety of natural language tasks while instruction and chat tuned models are intended for assistant-like chat.
-
-### Out-of-Scope Use
-
-<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
-
-Use in any manner that violates the license, any applicable laws or regluations, use in languages other than Romanian.
-
-
-
-## How to Get Started with the Model
-
-Use the code below to get started with the model.
-
-```python
-from transformers import AutoTokenizer, AutoModelForCausalLM
-
-tokenizer = AutoTokenizer.from_pretrained("OpenLLM-Ro/RoLlama3-8b-Instruct")
-model = AutoModelForCausalLM.from_pretrained("OpenLLM-Ro/RoLlama3-8b-Instruct")
-
-instruction = "Ce jocuri de societate pot juca cu prietenii mei?"
-chat = [
-        {"role": "system", "content": "Ești un asistent folositor, respectuos și onest. Încearcă să ajuți cât mai mult prin informațiile oferite, excluzând răspunsuri toxice, rasiste, sexiste, periculoase și ilegale."},
-        {"role": "user", "content": instruction},
-        ]
-prompt = tokenizer.apply_chat_template(chat, tokenize=False, system_message="")
-
-inputs = tokenizer.encode(prompt, add_special_tokens=False, return_tensors="pt")
-outputs = model.generate(input_ids=inputs, max_new_tokens=128)
-print(tokenizer.decode(outputs[0]))
-```
-
-## Academic Benchmarks
-
-<table>
-<tbody>
-<tr>
-<td><strong>Model</strong></td>
-<td><strong><center>Average</center></strong></td>
-<td><strong><center>ARC</center></strong></td>
-<td><strong><center>MMLU</center></strong></td>
-<td><strong><center>Winogrande</center></strong></td>
-<td><strong><center>Hellaswag</center></strong></td>
-<td><strong><center>GSM8k</center></strong></td>
-<td><strong><center>TruthfulQA</center></strong></td>
-</tr>
-<tr>
-<td>Llama-3-8B-Instruct</td><td><center>50.62</center></td><td><center>43.69</center></td><td><center>52.04</center></td><td><center>59.33</center></td><td><center>53.19</center></td><td><center><strong>43.87</strong></center></td><td><center><strong>51.59</strong></center></td>
-</tr>
-<tr>
-<td>RoLlama3-8b-Instruct-2024-06-28</td><td><center>50.56</center></td><td><center>44.70</center></td><td><center>52.19</center></td><td><center><strong>67.23</strong></center></td><td><center>57.69</center></td><td><center>30.23</center></td><td><center>51.34</center></td>
-</tr>
-<tr>
-<td><em>RoLlama3-8b-Instruct-2024-10-09</em></td><td><center><em><strong>52.21</strong></em></center></td><td><center><em><strong>47.94</strong></em></center></td><td><center><em><strong>53.50</strong></em></center></td><td><center><em>66.06</em></center></td><td><center><em><strong>59.72</strong></em></center></td><td><center><em>40.16</em></center></td><td><center><em>45.90</em></center></td>
-</tr>
-<tr>
-<td>RoLlama3-8b-Instruct-DPO-2024-10-09</td><td><center>49.96</center></td><td><center>46.29</center></td><td><center>53.29</center></td><td><center>65.57</center></td><td><center>58.15</center></td><td><center>34.77</center></td><td><center>41.70</center></td>
-</tr>
-</tbody>
-</table>
-
-  
-## Downstream tasks
-
-<table>
-<tbody>
-<tr>
-<td></td>
-<td colspan="4"><center><strong>LaRoSeDa</strong></center></td>
-<td colspan="4"><center><strong>WMT</strong></center></td>
-</tr>
-<tr>
-<td></td>
-<td colspan="2"><center><strong>Few-shot</strong></center></td>
-<td colspan="2"><center><strong>Finetuned</strong></center></td>
-<td colspan="2"><center><strong>Few-shot</strong></center></td>
-<td colspan="2"><center><strong>Finetuned</strong></center></td>
-</tr>
-<tr>
-<td><strong>Model</strong></td>
-<td><center><strong>Binary<br>(Macro F1)</strong></center></td>
-<td><center><strong>Multiclass<br>(Macro F1)</strong></center></td>
-<td><center><strong>Binary<br>(Macro F1)</strong></center></td>
-<td><center><strong>Multiclass<br>(Macro F1)</strong></center></td>
-<td><center><strong>EN-RO<br>(Bleu)</strong></center></td>
-<td><center><strong>RO-EN<br>(Bleu)</strong></center></td>
-<td><center><strong>EN-RO<br>(Bleu)</strong></center></td>
-<td><center><strong>RO-EN<br>(Bleu)</strong></center>
-</tr>
-<tr>
-<td>Llama-3-8B-Instruct</td><td><center>95.88</center></td><td><center>56.21</center></td><td><center><strong>98.53</strong></center></td><td><center>86.19</center></td><td><center>18.88</center></td><td><center><strong>30.98</strong></center></td><td><center><strong>28.02</strong></center></td><td><center>40.28</center></td>
-</tr>
-<tr>
-<td>RoLlama3-8b-Instruct-2024-06-28</td><td><center><strong>97.52</strong></center></td><td><center><strong>67.41</strong></center></td><td><center>94.15</center></td><td><center>87.13</center></td><td><center><strong>24.01</strong></center></td><td><center>27.36</center></td><td><center>26.53</center></td><td><center>40.36</center></td>
-</tr>
-<tr>
-<td><em>RoLlama3-8b-Instruct-2024-10-09</em></td><td><center><em>95.58</em></center></td><td><center><em>61.20</em></center></td><td><center><em>96.46</em></center></td><td><center><em><strong>87.26</strong></em></center></td><td><center><em>22.92</em></center></td><td><center><em>24.28</em></center></td><td><center><em>27.31</em></center></td><td><center><em><strong>40.52</strong></em></center></td>
-</tr>
-<tr>
-<td>RoLlama3-8b-Instruct-DPO-2024-10-09</td><td><center>97.48</center></td><td><center>54.00</center></td><td><center>-</center></td><td><center>-</center></td><td><center>22.09</center></td><td><center>23.00</center></td><td><center>-</center></td><td><center>-</center></td>
-</tr>
-</tbody>
-</table>
-
-
-<table>
-<tbody>
-<tr>
-<td></td>
-<td colspan="4"><center><strong>XQuAD</strong></center></td>
-<td colspan="4"><center><strong>STS</strong></center></td>
-</tr>
-<tr>
-<td></td>
-<td colspan="2"><center><strong>Few-shot</strong></center></td>
-<td colspan="2"><center><strong>Finetuned</strong></center></td>
-<td colspan="2"><center><strong>Few-shot</strong></center></td>
-<td colspan="2"><center><strong>Finetuned</strong></center></td>
-</tr>
-<tr>
-<td><strong>Model</strong></td>
-<td><center><strong>(EM)</strong></center></td>
-<td><center><strong>(F1)</strong></center></td>
-<td><center><strong>(EM)</strong></center></td>
-<td><center><strong>(F1)</strong></center></td>
-<td><center><strong>(Spearman)</strong></center></td>
-<td><center><strong>(Pearson)</strong></center></td>
-<td><center><strong>(Spearman)</strong></center></td>
-<td><center><strong>(Pearson)</strong></center></td>
-</tr>
-<tr>
-<td>Llama-3-8B-Instruct</td><td><center><strong>39.47</strong></center></td><td><center>58.67</center></td><td><center><strong>67.65</strong></center></td><td><center><strong>82.77</strong></center></td><td><center>73.04</center></td><td><center>72.36</center></td><td><center>83.49</center></td><td><center>84.06</center></td>
-</tr>
-<tr>
-<td>RoLlama3-8b-Instruct-2024-06-28</td><td><center>39.43</center></td><td><center><strong>59.50</strong></center></td><td><center>44.45</center></td><td><center>59.76</center></td><td><center>77.20</center></td><td><center>77.87</center></td><td><center>85.80</center></td><td><center>86.05</center></td>
-</tr>
-<tr>
-<td><em>RoLlama3-8b-Instruct-2024-10-09</em></td><td><center><em>18.89</em></center></td><td><center><em>31.79</em></center></td><td><center><em>50.84</em></center></td><td><center><em>65.18</em></center></td><td><center><em>77.60</em></center></td><td><center><em>76.86</em></center></td><td><center><em><strong>86.70</strong></em></center></td><td><center><em><strong>87.09</strong></em></center></td>
-</tr>
-<tr>
-<td>RoLlama3-8b-Instruct-DPO-2024-10-09</td><td><center>26.05</center></td><td><center>42.77</center></td><td><center>-</center></td><td><center>-</center></td><td><center><strong>79.64</strong></center></td><td><center><strong>79.52</strong></center></td><td><center>-</center></td><td><center>-</center></td>
-</tr>
-</tbody>
-</table>
-
-## MT-Bench
-
-<table>
-<tbody>
-<tr>
-<td><strong>Model</strong></td>
-<td><strong><center>Average</center></strong></td>
-<td><strong><center>1st turn</center></strong></td>
-<td><strong><center>2nd turn</center></strong></td>
-<td><strong><center>Answers in Ro</center></strong></td>
-</tr>
-<tr>
-<td>Llama-3-8B-Instruct</td><td><center><strong>5.96</strong></center></td><td><center>6.16</center></td><td><center><strong>5.76</strong></center></td><td><center>158/160</center></td>
-</tr>
-<tr>
-<td>RoLlama3-8b-Instruct-2024-06-28</td><td><center>5.15</center></td><td><center>6.03</center></td><td><center>4.28</center></td><td><center><strong>160/160</strong></center></td>
-</tr>
-<tr>
-<td><em>RoLlama3-8b-Instruct-2024-10-09</em></td><td><center><em>5.38</em></center></td><td><center><em>6.09</em></center></td><td><center><em>4.67</em></center></td><td><center><em><strong>160/160</strong></em></center></td>
-</tr>
-<tr>
-<td>RoLlama3-8b-Instruct-DPO-2024-10-09</td><td><center>5.87</center></td><td><center><strong>6.22</strong></center></td><td><center>5.49</center></td><td><center><strong>160/160</strong></center></td>
-</tr>
-</tbody>
-</table>
-
-
-## RoCulturaBench
-
-<table>
-<tbody>
-<tr>
-<td><strong>Model</strong></td>
-<td><strong><center>Average</center></strong></td>
-<td><strong><center>Answers in Ro</center></strong></td>
-</tr>
-<tr>
-<td>Llama-3-8B-Instruct</td><td><center><strong>4.62</strong></center></td><td><center><strong>100/100</strong></center></td>
-</tr>
-<tr>
-<td>RoLlama3-8b-Instruct-2024-06-28</td><td><center>3.71</center></td><td><center><strong>100/100</strong></center></td>
-</tr>
-<tr>
-<td><em>RoLlama3-8b-Instruct-2024-10-09</em></td><td><center><em>3.81</em></center></td><td><center><em><strong>100/100</strong></em></center></td>
-</tr>
-<tr>
-<td>RoLlama3-8b-Instruct-DPO-2024-10-09</td><td><center>4.40</center></td><td><center><strong>100/100</strong></center></td>
-</tr>
-</tbody>
-</table>
-
-
-
-## RoLlama3 Model Family
-
-| Model              | Link  |
-|--------------------|:--------:|
-|RoLlama3-8b-Instruct-2024-06-28| [link](https://huggingface.co/OpenLLM-Ro/RoLlama3-8b-Instruct-2024-06-28) |
-|*RoLlama3-8b-Instruct-2024-10-09*| [link](https://huggingface.co/OpenLLM-Ro/RoLlama3-8b-Instruct-2024-10-09) |
-|RoLlama3-8b-Instruct-DPO-2024-10-09| [link](https://huggingface.co/OpenLLM-Ro/RoLlama3-8b-Instruct-DPO-2024-10-09) |
-
-
-## Citation 
-
-```
-@misc{masala2024vorbecstiromanecsterecipetrain,
-      title={"Vorbe\c{s}ti Rom\^ane\c{s}te?" A Recipe to Train Powerful Romanian LLMs with English Instructions}, 
-      author={Mihai Masala and Denis C. Ilie-Ablachim and Alexandru Dima and Dragos Corlatescu and Miruna Zavelca and Ovio Olaru and Simina Terian-Dan and Andrei Terian-Dan and Marius Leordeanu and Horia Velicu and Marius Popescu and Mihai Dascalu and Traian Rebedea},
-      year={2024},
-      eprint={2406.18266},
-      archivePrefix={arXiv},
-      primaryClass={cs.CL},
-      url={https://arxiv.org/abs/2406.18266}, 
-}
-```
-<!-- **APA:**
-
+---
+license: cc-by-nc-4.0
+language:
+- ro
+base_model:
+- meta-llama/Meta-Llama-3-8B-Instruct
+datasets:
+- OpenLLM-Ro/ro_sft_alpaca
+- OpenLLM-Ro/ro_sft_alpaca_gpt4
+- OpenLLM-Ro/ro_sft_dolly
+- OpenLLM-Ro/ro_sft_selfinstruct_gpt4
+- OpenLLM-Ro/ro_sft_norobots
+- OpenLLM-Ro/ro_sft_orca
+- OpenLLM-Ro/ro_sft_camel
+- OpenLLM-Ro/ro_sft_oasst
+- OpenLLM-Ro/ro_sft_ultrachat
+- OpenLLM-Ro/ro_sft_magpie_mt
+- OpenLLM-Ro/ro_sft_magpie_reasoning
+model-index:
+    - name: OpenLLM-Ro/RoLlama3-8b-Instruct-2025-04-23
+      results:
+        - task:
+            type: text-generation
+          dataset:
+            name: RoMT-Bench
+            type: RoMT-Bench
+          metrics:
+            - name: Score
+              type: Score
+              value: 6.39
+        - task:
+            type: text-generation
+          dataset:
+            name: RoCulturaBench
+            type: RoCulturaBench
+          metrics:
+            - name: Score
+              type: Score
+              value: 4.05
+        - task:
+            type: text-generation
+          dataset:
+            name: Romanian_Academic_Benchmarks
+            type: Romanian_Academic_Benchmarks
+          metrics:
+            - name: Average accuracy
+              type: accuracy
+              value: 54.66
+        - task:
+            type: text-generation
+          dataset:
+            name: OpenLLM-Ro/ro_arc_challenge
+            type: OpenLLM-Ro/ro_arc_challenge
+          metrics:
+            - name: Average accuracy
+              type: accuracy
+              value: 50.31
+        - task:
+            type: text-generation
+          dataset:
+            name: OpenLLM-Ro/ro_mmlu
+            type: OpenLLM-Ro/ro_mmlu
+          metrics:
+            - name: Average accuracy
+              type: accuracy
+              value: 55.91
+        - task:
+            type: text-generation
+          dataset:
+            name: OpenLLM-Ro/ro_winogrande
+            type: OpenLLM-Ro/ro_winogrande
+          metrics:
+            - name: Average accuracy
+              type: accuracy
+              value: 67.01
+        - task:
+            type: text-generation
+          dataset:
+            name: OpenLLM-Ro/ro_hellaswag
+            type: OpenLLM-Ro/ro_hellaswag
+          metrics:
+            - name: Average accuracy
+              type: accuracy
+              value: 61.73
+        - task:
+            type: text-generation
+          dataset:
+            name: OpenLLM-Ro/ro_gsm8k
+            type: OpenLLM-Ro/ro_gsm8k
+          metrics:
+            - name: Average accuracy
+              type: accuracy
+              value: 47.41
+        - task:
+            type: text-generation
+          dataset:
+            name: OpenLLM-Ro/ro_truthfulqa
+            type: OpenLLM-Ro/ro_truthfulqa
+          metrics:
+            - name: Average accuracy
+              type: accuracy
+              value: 45.61
+        - task:
+            type: text-generation
+          dataset:
+            name: LaRoSeDa_binary
+            type: LaRoSeDa_binary
+          metrics:
+            - name: Average macro-f1
+              type: macro-f1
+              value: 96.21
+        - task:
+            type: text-generation
+          dataset:
+            name: LaRoSeDa_multiclass
+            type: LaRoSeDa_multiclass
+          metrics:
+            - name: Average macro-f1
+              type: macro-f1
+              value: 59.15
+        - task:
+            type: text-generation
+          dataset:
+            name: WMT_EN-RO
+            type: WMT_EN-RO
+          metrics:
+            - name: Average bleu
+              type: bleu
+              value: 23.32
+        - task:
+            type: text-generation
+          dataset:
+            name: WMT_RO-EN
+            type: WMT_RO-EN
+          metrics:
+            - name: Average bleu
+              type: bleu
+              value: 22.50
+        - task:
+            type: text-generation
+          dataset:
+            name: XQuAD
+            type: XQuAD
+          metrics:
+            - name: Average exact_match
+              type: exact_match
+              value: 11.01
+        - task:
+            type: text-generation
+          dataset:
+            name: XQuAD
+            type: XQuAD
+          metrics:
+            - name: Average f1
+              type: f1
+              value: 23.55
+        - task:
+            type: text-generation
+          dataset:
+            name: STS
+            type: STS
+          metrics:
+            - name: Average spearman
+              type: spearman
+              value: 76.78
+        - task:
+            type: text-generation
+          dataset:
+            name: STS
+            type: STS
+          metrics:
+            - name: Average pearson
+              type: pearson
+              value: 74.36
+        - task:
+            type: text-generation
+          dataset:
+            name: RoMT-Bench
+            type: RoMT-Bench
+          metrics:
+            - name: First turn
+              type: Score
+              value: 7.12
+            - name: Second turn
+              type: Score
+              value: 5.66
+        - task:
+            type: text-generation
+          dataset:
+            name: OpenLLM-Ro/ro_arc_challenge
+            type: OpenLLM-Ro/ro_arc_challenge
+          metrics:
+            - name: 0-shot 
+              type: accuracy
+              value: 48.33
+            - name: 1-shot 
+              type: accuracy
+              value: 49.27
+            - name: 3-shot 
+              type: accuracy
+              value: 49.19
+            - name: 5-shot 
+              type: accuracy
+              value: 50.90
+            - name: 10-shot 
+              type: accuracy
+              value: 51.67
+            - name: 25-shot 
+              type: accuracy
+              value: 52.53
+        - task:
+            type: text-generation
+          dataset:
+            name: OpenLLM-Ro/ro_mmlu
+            type: OpenLLM-Ro/ro_mmlu
+          metrics:
+            - name: 0-shot 
+              type: accuracy
+              value: 54.17
+            - name: 1-shot 
+              type: accuracy
+              value: 56.19
+            - name: 3-shot 
+              type: accuracy
+              value: 56.90
+            - name: 5-shot 
+              type: accuracy
+              value: 56.37
+        - task:
+            type: text-generation
+          dataset:
+            name: OpenLLM-Ro/ro_winogrande
+            type: OpenLLM-Ro/ro_winogrande
+          metrics:
+            - name: 0-shot 
+              type: accuracy
+              value: 65.82
+            - name: 1-shot 
+              type: accuracy
+              value: 66.22
+            - name: 3-shot 
+              type: accuracy
+              value: 66.85
+            - name: 5-shot 
+              type: accuracy
+              value: 69.14
+        - task:
+            type: text-generation
+          dataset:
+            name: OpenLLM-Ro/ro_hellaswag
+            type: OpenLLM-Ro/ro_hellaswag
+          metrics:
+            - name: 0-shot 
+              type: accuracy
+              value: 61.67
+            - name: 1-shot 
+              type: accuracy
+              value: 62.06
+            - name: 3-shot 
+              type: accuracy
+              value: 61.73
+            - name: 5-shot 
+              type: accuracy
+              value: 61.28
+            - name: 10-shot 
+              type: accuracy
+              value: 61.93
+        - task:
+            type: text-generation
+          dataset:
+            name: OpenLLM-Ro/ro_gsm8k
+            type: OpenLLM-Ro/ro_gsm8k
+          metrics:
+            - name: 1-shot 
+              type: accuracy
+              value: 35.63
+            - name: 3-shot 
+              type: accuracy
+              value: 51.33
+            - name: 5-shot 
+              type: accuracy
+              value: 55.27
+        - task:
+            type: text-generation
+          dataset:
+            name: LaRoSeDa_binary
+            type: LaRoSeDa_binary
+          metrics:
+            - name: 0-shot 
+              type: macro-f1
+              value: 94.05
+            - name: 1-shot 
+              type: macro-f1
+              value: 96.46
+            - name: 3-shot 
+              type: macro-f1
+              value: 96.97
+            - name: 5-shot 
+              type: macro-f1
+              value: 97.37
+        - task:
+            type: text-generation
+          dataset:
+            name: LaRoSeDa_multiclass
+            type: LaRoSeDa_multiclass
+          metrics:
+            - name: 0-shot 
+              type: macro-f1
+              value: 60.34
+            - name: 1-shot 
+              type: macro-f1
+              value: 60.94
+            - name: 3-shot 
+              type: macro-f1
+              value: 54.55
+            - name: 5-shot 
+              type: macro-f1
+              value: 60.77
+        - task:
+            type: text-generation
+          dataset:
+            name: WMT_EN-RO
+            type: WMT_EN-RO
+          metrics:
+            - name: 0-shot 
+              type: bleu
+              value: 5.38
+            - name: 1-shot 
+              type: bleu
+              value: 29.60
+            - name: 3-shot 
+              type: bleu
+              value: 30.62
+            - name: 5-shot 
+              type: bleu
+              value: 27.67
+        - task:
+            type: text-generation
+          dataset:
+            name: WMT_RO-EN
+            type: WMT_RO-EN
+          metrics:
+            - name: 0-shot 
+              type: bleu
+              value: 1.14
+            - name: 1-shot 
+              type: bleu
+              value: 19.96
+            - name: 3-shot 
+              type: bleu
+              value: 34.22
+            - name: 5-shot 
+              type: bleu
+              value: 34.69
+        - task:
+            type: text-generation
+          dataset:
+            name: XQuAD_EM
+            type: XQuAD_EM
+          metrics:
+            - name: 0-shot 
+              type: exact_match
+              value: 16.39
+            - name: 1-shot 
+              type: exact_match
+              value: 18.49
+            - name: 3-shot 
+              type: exact_match
+              value: 5.46
+            - name: 5-shot 
+              type: exact_match
+              value: 3.70
+        - task:
+            type: text-generation
+          dataset:
+            name: XQuAD_F1
+            type: XQuAD_F1
+          metrics:
+            - name: 0-shot 
+              type: f1
+              value: 33.84
+            - name: 1-shot 
+              type: f1
+              value: 29.11
+            - name: 3-shot 
+              type: f1
+              value: 15.27
+            - name: 5-shot 
+              type: f1
+              value: 15.97
+        - task:
+            type: text-generation
+          dataset:
+            name: STS_Spearman
+            type: STS_Spearman
+          metrics:
+            - name: 1-shot 
+              type: spearman
+              value: 76.64
+            - name: 3-shot 
+              type: spearman
+              value: 76.88
+            - name: 5-shot 
+              type: spearman
+              value: 76.82
+        - task:
+            type: text-generation
+          dataset:
+            name: STS_Pearson
+            type: STS_Pearson
+          metrics:
+            - name: 1-shot 
+              type: pearson
+              value: 73.14
+            - name: 3-shot 
+              type: pearson
+              value: 74.78
+            - name: 5-shot 
+              type: pearson
+              value: 75.16
+
+---
+
+# Model Card for Model ID
+
+*Built with Meta Llama 3*
+
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+RoLlama3 is a family of pretrained and fine-tuned generative text models for Romanian. This is the repository for the **instruct 8B model**. Links to other models can be found at the bottom of this page.
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+OpenLLM-Ro represents the first open-source effort to build a LLM specialized for Romanian. OpenLLM-Ro developed and publicly releases a collection of Romanian LLMs, both in the form of foundational model and instruct and chat variants.
+
+
+- **Developed by:** OpenLLM-Ro
+<!-- - **Funded by [optional]:** [More Information Needed] -->
+<!-- - **Shared by [optional]:** [More Information Needed] -->
+<!-- - **Model type:** [More Information Needed] -->
+- **Language(s):** Romanian
+- **License:** cc-by-nc-4.0
+- **Finetuned from model:** [Meta-Llama-3-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct)
+- **Trained using:** [RoAlpaca](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_alpaca), [RoAlpacaGPT4](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_alpaca_gpt4), [RoDolly](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_dolly), [RoSelfInstruct](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_selfinstruct_gpt4), [RoNoRobots](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_norobots), [RoOrca](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_orca), [RoCamel](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_camel), [RoOpenAssistant](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_oasst), [RoUltraChat](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_ultrachat), [RoMagpiePro](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_magpie_mt), [RoMagpieReasoning](https://huggingface.co/datasets/OpenLLM-Ro/ro_sft_magpie_reasoning)
+
+
+### Model Sources
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** https://github.com/OpenLLM-Ro/LLaMA-Factory
+- **Paper:** https://arxiv.org/abs/2406.18266
+
+## Intended Use
+
+### Intended Use Cases
+
+RoLlama3 is intented for research use in Romanian. Base models can be adapted for a variety of natural language tasks while instruction and chat tuned models are intended for assistant-like chat.
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+Use in any manner that violates the license, any applicable laws or regluations, use in languages other than Romanian.
+
+
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+```python
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
+tokenizer = AutoTokenizer.from_pretrained("OpenLLM-Ro/RoLlama3-8b-Instruct-2025-04-23")
+model = AutoModelForCausalLM.from_pretrained("OpenLLM-Ro/RoLlama3-8b-Instruct-2025-04-23")
+
+instruction = "Ce jocuri de societate pot juca cu prietenii mei?"
+chat = [
+        {"role": "system", "content": "Ești un asistent folositor, respectuos și onest. Încearcă să ajuți cât mai mult prin informațiile oferite, excluzând răspunsuri toxice, rasiste, sexiste, periculoase și ilegale."},
+        {"role": "user", "content": instruction},
+        ]
+prompt = tokenizer.apply_chat_template(chat, tokenize=False, system_message="")
+
+inputs = tokenizer.encode(prompt, add_special_tokens=False, return_tensors="pt")
+outputs = model.generate(input_ids=inputs, max_new_tokens=128)
+print(tokenizer.decode(outputs[0]))
+```
+
+## Academic Benchmarks
+
+<table>
+<tbody>
+<tr>
+<td><strong>Model</strong></td>
+<td><strong><center>Average</center></strong></td>
+<td><strong><center>ARC</center></strong></td>
+<td><strong><center>MMLU</center></strong></td>
+<td><strong><center>Winogrande</center></strong></td>
+<td><strong><center>Hellaswag</center></strong></td>
+<td><strong><center>GSM8k</center></strong></td>
+<td><strong><center>TruthfulQA</center></strong></td>
+</tr>
+<tr>
+<td>Llama-3-8B-Instruct</td><td><center>50.62</center></td><td><center>43.69</center></td><td><center>52.04</center></td><td><center>59.33</center></td><td><center>53.19</center></td><td><center>43.87</center></td><td><center>51.59</center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-2024-06-28</td><td><center>50.56</center></td><td><center>44.70</center></td><td><center>52.19</center></td><td><center><strong>67.23</strong></center></td><td><center>57.69</center></td><td><center>30.23</center></td><td><center>51.34</center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-2024-10-09</td><td><center>52.21</center></td><td><center>47.94</center></td><td><center>53.50</center></td><td><center>66.06</center></td><td><center>59.72</center></td><td><center>40.16</center></td><td><center>45.90</center></td>
+</tr>
+<tr>
+<td><em>RoLlama3-8b-Instruct-2025-04-23</em></td><td><center><em>54.66</em></center></td><td><center><em>50.31</em></center></td><td><center><em><strong>55.91</strong></em></center></td><td><center><em>67.01</em></center></td><td><center><em><strong>61.73</strong></em></center></td><td><center><em><strong>47.41</strong></em></center></td><td><center><em>45.61</em></center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-DPO-2024-10-09</td><td><center>49.96</center></td><td><center>46.29</center></td><td><center>53.29</center></td><td><center>65.57</center></td><td><center>58.15</center></td><td><center>34.77</center></td><td><center>41.70</center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-DPO-2025-04-23</td><td><center><strong>55.86</strong></center></td><td><center><strong>52.26</strong></center></td><td><center>55.35</center></td><td><center>66.62</center></td><td><center>59.93</center></td><td><center>43.95</center></td><td><center><strong>57.06</strong></center></td>
+</tr>
+</tbody>
+</table>
+
+  
+## Downstream tasks
+
+<table>
+<tbody>
+<tr>
+<td></td>
+<td colspan="4"><center><strong>LaRoSeDa</strong></center></td>
+<td colspan="4"><center><strong>WMT</strong></center></td>
+</tr>
+<tr>
+<td></td>
+<td colspan="2"><center><strong>Few-shot</strong></center></td>
+<td colspan="2"><center><strong>Finetuned</strong></center></td>
+<td colspan="2"><center><strong>Few-shot</strong></center></td>
+<td colspan="2"><center><strong>Finetuned</strong></center></td>
+</tr>
+<tr>
+<td><strong>Model</strong></td>
+<td><center><strong>Binary<br>(Macro F1)</strong></center></td>
+<td><center><strong>Multiclass<br>(Macro F1)</strong></center></td>
+<td><center><strong>Binary<br>(Macro F1)</strong></center></td>
+<td><center><strong>Multiclass<br>(Macro F1)</strong></center></td>
+<td><center><strong>EN-RO<br>(Bleu)</strong></center></td>
+<td><center><strong>RO-EN<br>(Bleu)</strong></center></td>
+<td><center><strong>EN-RO<br>(Bleu)</strong></center></td>
+<td><center><strong>RO-EN<br>(Bleu)</strong></center>
+</tr>
+<tr>
+<td>Llama-3-8B-Instruct</td><td><center>95.88</center></td><td><center>56.21</center></td><td><center><strong>98.53</strong></center></td><td><center>86.19</center></td><td><center>18.88</center></td><td><center><strong>30.98</strong></center></td><td><center><strong>28.02</strong></center></td><td><center>40.28</center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-2024-06-28</td><td><center>97.52</center></td><td><center><strong>67.41</strong></center></td><td><center>94.15</center></td><td><center>87.13</center></td><td><center><strong>24.01</strong></center></td><td><center>27.36</center></td><td><center>26.53</center></td><td><center>40.36</center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-2024-10-09</td><td><center>95.58</center></td><td><center>61.20</center></td><td><center>96.46</center></td><td><center><strong>87.26</strong></center></td><td><center>22.92</center></td><td><center>24.28</center></td><td><center>27.31</center></td><td><center><strong>40.52</strong></center></td>
+</tr>
+<tr>
+<td><em>RoLlama3-8b-Instruct-2025-04-23</em></td><td><center><em>96.21</em></center></td><td><center><em>59.15</em></center></td><td><center><em>-</em></center></td><td><center><em>-</em></center></td><td><center><em>23.32</em></center></td><td><center><em>22.50</em></center></td><td><center><em>-</em></center></td><td><center><em>-</em></center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-DPO-2024-10-09</td><td><center>97.48</center></td><td><center>54.00</center></td><td><center>-</center></td><td><center>-</center></td><td><center>22.09</center></td><td><center>23.00</center></td><td><center>-</center></td><td><center>-</center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-DPO-2025-04-23</td><td><center><strong>97.60</strong></center></td><td><center>62.16</center></td><td><center>-</center></td><td><center>-</center></td><td><center>18.14</center></td><td><center>14.13</center></td><td><center>-</center></td><td><center>-</center></td>
+</tr>
+</tbody>
+</table>
+
+
+<table>
+<tbody>
+<tr>
+<td></td>
+<td colspan="4"><center><strong>XQuAD</strong></center></td>
+<td colspan="4"><center><strong>STS</strong></center></td>
+</tr>
+<tr>
+<td></td>
+<td colspan="2"><center><strong>Few-shot</strong></center></td>
+<td colspan="2"><center><strong>Finetuned</strong></center></td>
+<td colspan="2"><center><strong>Few-shot</strong></center></td>
+<td colspan="2"><center><strong>Finetuned</strong></center></td>
+</tr>
+<tr>
+<td><strong>Model</strong></td>
+<td><center><strong>(EM)</strong></center></td>
+<td><center><strong>(F1)</strong></center></td>
+<td><center><strong>(EM)</strong></center></td>
+<td><center><strong>(F1)</strong></center></td>
+<td><center><strong>(Spearman)</strong></center></td>
+<td><center><strong>(Pearson)</strong></center></td>
+<td><center><strong>(Spearman)</strong></center></td>
+<td><center><strong>(Pearson)</strong></center></td>
+</tr>
+<tr>
+<td>Llama-3-8B-Instruct</td><td><center><strong>39.47</strong></center></td><td><center>58.67</center></td><td><center><strong>67.65</strong></center></td><td><center><strong>82.77</strong></center></td><td><center>73.04</center></td><td><center>72.36</center></td><td><center>83.49</center></td><td><center>84.06</center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-2024-06-28</td><td><center>39.43</center></td><td><center><strong>59.50</strong></center></td><td><center>44.45</center></td><td><center>59.76</center></td><td><center>77.20</center></td><td><center>77.87</center></td><td><center>85.80</center></td><td><center>86.05</center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-2024-10-09</td><td><center>18.89</center></td><td><center>31.79</center></td><td><center>50.84</center></td><td><center>65.18</center></td><td><center>77.60</center></td><td><center>76.86</center></td><td><center><strong>86.70</strong></center></td><td><center><strong>87.09</strong></center></td>
+</tr>
+<tr>
+<td><em>RoLlama3-8b-Instruct-2025-04-23</em></td><td><center><em>11.01</em></center></td><td><center><em>23.55</em></center></td><td><center><em>-</em></center></td><td><center><em>-</em></center></td><td><center><em>76.78</em></center></td><td><center><em>74.36</em></center></td><td><center><em>-</em></center></td><td><center><em>-</em></center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-DPO-2024-10-09</td><td><center>26.05</center></td><td><center>42.77</center></td><td><center>-</center></td><td><center>-</center></td><td><center><strong>79.64</strong></center></td><td><center><strong>79.52</strong></center></td><td><center>-</center></td><td><center>-</center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-DPO-2025-04-23</td><td><center>30.65</center></td><td><center>46.29</center></td><td><center>-</center></td><td><center>-</center></td><td><center>67.62</center></td><td><center>67.82</center></td><td><center>-</center></td><td><center>-</center></td>
+</tr>
+</tbody>
+</table>
+
+## MT-Bench
+
+<table>
+<tbody>
+<tr>
+<td><strong>Model</strong></td>
+<td><strong><center>Average</center></strong></td>
+<td><strong><center>1st turn</center></strong></td>
+<td><strong><center>2nd turn</center></strong></td>
+<td><strong><center>Answers in Ro</center></strong></td>
+</tr>
+<tr>
+<td>Llama-3-8B-Instruct</td><td><center>5.96</center></td><td><center>6.16</center></td><td><center>5.76</center></td><td><center>158/160</center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-2024-06-28</td><td><center>5.15</center></td><td><center>6.03</center></td><td><center>4.28</center></td><td><center><strong>160/160</strong></center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-2024-10-09</td><td><center>5.38</center></td><td><center>6.09</center></td><td><center>4.67</center></td><td><center><strong>160/160</strong></center></td>
+</tr>
+<tr>
+<td><em>RoLlama3-8b-Instruct-2025-04-23</em></td><td><center><em>6.39</em></center></td><td><center><em><strong>7.12</strong></em></center></td><td><center><em>5.66</em></center></td><td><center><em><strong>160/160</strong></em></center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-DPO-2024-10-09</td><td><center>5.87</center></td><td><center>6.22</center></td><td><center>5.49</center></td><td><center><strong>160/160</strong></center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-DPO-2025-04-23</td><td><center><strong>6.67</strong></center></td><td><center>6.81</center></td><td><center><strong>6.54</strong></center></td><td><center><strong>160/160</strong></center></td>
+</tr>
+</tbody>
+</table>
+
+
+## RoCulturaBench
+
+<table>
+<tbody>
+<tr>
+<td><strong>Model</strong></td>
+<td><strong><center>Average</center></strong></td>
+<td><strong><center>Answers in Ro</center></strong></td>
+</tr>
+<tr>
+<td>Llama-3-8B-Instruct</td><td><center>4.62</center></td><td><center><strong>100/100</strong></center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-2024-06-28</td><td><center>3.71</center></td><td><center><strong>100/100</strong></center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-2024-10-09</td><td><center>3.81</center></td><td><center><strong>100/100</strong></center></td>
+</tr>
+<tr>
+<td><em>RoLlama3-8b-Instruct-2025-04-23</em></td><td><center><em>4.05</em></center></td><td><center><em><strong>100/100</strong></em></center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-DPO-2024-10-09</td><td><center>4.40</center></td><td><center><strong>100/100</strong></center></td>
+</tr>
+<tr>
+<td>RoLlama3-8b-Instruct-DPO-2025-04-23</td><td><center><strong>4.83</strong></center></td><td><center><strong>100/100</strong></center></td>
+</tr>
+</tbody>
+</table>
+
+
+
+## RoLlama3 Model Family
+
+| Model              | Link  |
+|--------------------|:--------:|
+|RoLlama3-8b-Instruct-2024-06-28| [link](https://huggingface.co/OpenLLM-Ro/RoLlama3-8b-Instruct-2024-06-28) |
+|RoLlama3-8b-Instruct-2024-10-09| [link](https://huggingface.co/OpenLLM-Ro/RoLlama3-8b-Instruct-2024-10-09) |
+|*RoLlama3-8b-Instruct-2025-04-23*| [link](https://huggingface.co/OpenLLM-Ro/RoLlama3-8b-Instruct-2025-04-23) |
+|RoLlama3-8b-Instruct-DPO-2024-10-09| [link](https://huggingface.co/OpenLLM-Ro/RoLlama3-8b-Instruct-DPO-2024-10-09) |
+|RoLlama3-8b-Instruct-DPO-2025-04-23| [link](https://huggingface.co/OpenLLM-Ro/RoLlama3-8b-Instruct-DPO-2025-04-23) |
+
+
+## Citation 
+
+```
+@misc{masala2024vorbecstiromanecsterecipetrain,
+      title={"Vorbe\c{s}ti Rom\^ane\c{s}te?" A Recipe to Train Powerful Romanian LLMs with English Instructions}, 
+      author={Mihai Masala and Denis C. Ilie-Ablachim and Alexandru Dima and Dragos Corlatescu and Miruna Zavelca and Ovio Olaru and Simina Terian-Dan and Andrei Terian-Dan and Marius Leordeanu and Horia Velicu and Marius Popescu and Mihai Dascalu and Traian Rebedea},
+      year={2024},
+      eprint={2406.18266},
+      archivePrefix={arXiv},
+      primaryClass={cs.CL},
+      url={https://arxiv.org/abs/2406.18266}, 
+}
+```
+<!-- **APA:**
+
 [More Information Needed]  -->