jenbenarye commited on
Commit
a431110
Β·
1 Parent(s): becdb13

edited text

Browse files
Files changed (1) hide show
  1. app/app.py +16 -7
app/app.py CHANGED
@@ -737,15 +737,24 @@ with gr.Blocks(css=css, js=js) as demo:
737
 
738
  with gr.Accordion("About") as explanation:
739
  gr.Markdown(f"""
740
- FeeL is a collaboration between Hugging Face and MIT.
741
- It is a community-driven project to provide a real-time feedback loop for LMs, where your feedback is continuously used to fine-tune the underlying models.
742
- The [dataset](https://huggingface.co/datasets/{scheduler.repo_id}), [code](https://github.com/huggingface/feel) and [models](https://huggingface.co/collections/feel-fl/feel-models-67a9b6ef0fdd554315e295e8) are public.
743
 
744
- Chat with the model using text and images and provide feedback in different ways:
745
 
746
- - ✏️ Edit a message
747
- - πŸ‘/πŸ‘Ž Like or dislike a message
748
- - πŸ”„ Regenerate a message
 
 
 
 
 
 
 
 
 
 
 
749
 
750
  """)
751
 
 
737
 
738
  with gr.Accordion("About") as explanation:
739
  gr.Markdown(f"""
740
+ ## Welcome to FeeL: Improving LMs for All Languages
 
 
741
 
742
+ **FeeL** (Feedback Loop) is a community-driven project by MIT, Hugging Face and IBM that aims to make language models better in *all languages*.
743
 
744
+ ### Why This Matters
745
+ Have you ever tried using an AI in your native language only to get responses that barely make sense? Most AI improvements prioritize widely spoken languages, while others fall behind. FeeL changes this by letting YOU shape how models respond in your language.
746
+
747
+ ### What You Can Do
748
+ 1. **Select your language** from the dropdown menu (or add a new one if yours is missing)
749
+ 2. **Chat with the model** in your language
750
+ 3. **Provide feedback** on each response using:
751
+ - πŸ‘/πŸ‘Ž Like or dislike responses
752
+ - ✏️ Edit responses to sound more natural or correct
753
+ - πŸ”„ Regenerate to try another response
754
+
755
+ Your feedback is directly used to fine-tune the model in real-time. The more you interact, the better the model becomes for your language community.
756
+
757
+ All [data](https://huggingface.co/datasets/{scheduler.repo_id}), [code](https://github.com/huggingface/feel) and [models](https://huggingface.co/collections/feel-fl/feel-models-67a9b6ef0fdd554315e295e8) are publicly available for research and development.
758
 
759
  """)
760