| --- |
| title: GoTo AI Leaderboard |
| emoji: 🥇 |
| colorFrom: green |
| colorTo: indigo |
| sdk: gradio |
| app_file: app.py |
| pinned: true |
| license: apache-2.0 |
| short_description: GoTo-AI leaderboard |
| sdk_version: 5.43.1 |
| tags: |
| - leaderboard |
| --- |
| |
| # Start the configuration |
|
|
| Leaderboard text, benchmark descriptions, citations, and table structure are configured in `src/config.py`. |
|
|
| Model performance data is stored in `config/model_performance.jsonl`, one JSON object per line, e.g.: |
| ```json |
| { |
| "model_name": "org/model", |
| "model_type": "Instruct", |
| "model_size": "8B", |
| "id": { |
| "nlu": {"sentiment": {"total": 70.6}, "total": 68.0}, |
| "total": 46.1 |
| } |
| } |
| ``` |
| Each language key (`id`, `jv`, `su`, `ban`, `bbc`, `indommlu`) nests its competency/task scores, with a `total` at each level. |
|
|
| # Code logic for more complex edits |
|
|
| - `src/config.py` — base model info columns, per-language column/tab configuration, and all display text (title, intro, benchmark/score-calculation info, citations). |
| - `src/populate.py` — flattens `config/model_performance.jsonl` into the Overall table plus one table per language/benchmark tab. |
| - `app.py` — builds the Gradio UI: tabs by model type (Instruct/Base) and by language, plus informational accordions. |
|
|
