Files changed (2) hide show
  1. config/model_performance.jsonl +0 -0
  2. src/config.py +6 -5
config/model_performance.jsonl CHANGED
The diff for this file is too large to render. See raw diff
 
src/config.py CHANGED
@@ -5,7 +5,7 @@ TITLE = """<h1 align="center" id="space-title">GoTo-AI Leaderboard</h1>"""
5
 
6
  # Introduction text providing an overview of the leaderboard
7
  INTRODUCTION_TEXT = """
8
- GoTo-AI (Indonesian language for "close friends") is a collection of large language models which has been pretrained and instruct-tuned for Indonesian language and its various local languages.
9
  This leaderboard evaluates general language capabilities of GoTo-AI and other open source models using SEA-HELM and IndoMMLU, focusing on Indonesian, Javanese, Sundanese, Balinese, and Batak.
10
  """
11
 
@@ -48,14 +48,15 @@ IndoMMLU covers various subjects and educational levels, including STEM, social
48
 
49
  # Explanation of score calculation methodology
50
  INFO_SCORE_CALCULATION = """
51
- - The **overall score** for a language is computed as the **average** of all competency scores.
52
- - Each **competency score** is computed as the **average** of its tasks.
53
- - Normalization is applied for classification tasks by substracting the random baseline score and scaling it to the range of 0-100.
 
54
  """
55
 
56
  # Placeholder information about GoTo and GoTo AI
57
  INFO_GOTO_AI = """
58
- GoTo-AI (Indonesian language for “close friends”) is a local open source Large Language Model (LLM) ecosystem in Indonesian language, co-initiated by Indonesian tech and telecommunication companies: GoTo Group and Indosat Ooredoo Hutchison. GoTo-AI ecosystem aims to empower Indonesians who want to develop AI-based services and applications using Bahasa Indonesia and its various local languages.
59
 
60
  We are supported by research centers and global tech experts such as AI Singapore to train the model to gain general language understanding.
61
 
 
5
 
6
  # Introduction text providing an overview of the leaderboard
7
  INTRODUCTION_TEXT = """
8
+ GoTo-AI is a collection of large language models which has been pretrained and instruct-tuned for Indonesian language and its various local languages.
9
  This leaderboard evaluates general language capabilities of GoTo-AI and other open source models using SEA-HELM and IndoMMLU, focusing on Indonesian, Javanese, Sundanese, Balinese, and Batak.
10
  """
11
 
 
48
 
49
  # Explanation of score calculation methodology
50
  INFO_SCORE_CALCULATION = """
51
+ - The evaluation runs 8 separate times to capture variations in non-deterministic outputs.
52
+ - 30 bootstrap samples are randomly drawn with replacement from the pooled evaluation data and the scores are averaged to a final score.
53
+ - The overall score for a language is computed as the average of all competency scores. Each competency score is computed as the average of its tasks.
54
+ - Normalization is applied for multi-choice task by substracting the random baseline score and scaling it to the range of 0-100.
55
  """
56
 
57
  # Placeholder information about GoTo and GoTo AI
58
  INFO_GOTO_AI = """
59
+ GoTo-AI is a local open source Large Language Model (LLM) ecosystem in Indonesian language, co-initiated by Indonesian tech and telecommunication companies: GoTo Group and Indosat Ooredoo Hutchison. GoTo-AI ecosystem aims to empower Indonesians who want to develop AI-based services and applications using Bahasa Indonesia and its various local languages.
60
 
61
  We are supported by research centers and global tech experts such as AI Singapore to train the model to gain general language understanding.
62