Spaces:
Sleeping
Sleeping
Commit
·
9887edb
1
Parent(s):
c905925
Update short description for book blurbs
Browse files
app.py
CHANGED
|
@@ -160,7 +160,7 @@ def log_blurb_and_vote(prompt, blurb, vote, user_info: gr.OAuthProfile | None, *
|
|
| 160 |
return f"Logged: {vote} by user {user_id}", gr.Row.update(visible=False)
|
| 161 |
|
| 162 |
|
| 163 |
-
short_description = """Vote on book blurbs generated by
|
| 164 |
"""
|
| 165 |
|
| 166 |
full_description = """Large Language Models are already strong assistants for technical tasks like coding. \n\nIncreasingly, they are also being used to help with tasks like copywriting. The jury is out on whether the texts produced by language models in these applications are very appealing; "let\'s delve into" is a common example of the clunky cliche-ridden text that LLMs often produce. \n\n<br> However, there is growing interest in using LLMs to help with more creative tasks. Outside of larger companies, a growing community of people fine-tuning LLMs for all kinds of creative tasks. \n\nSome writers want to use LLMs – not as a replacement but as a companion – in their writing process. <br> One of the requirements for building models which are better able to generate responses people like is having preference data. Preference data come in many forms but essentially boil to a dataset that contains some kind of signal for whether people like or dislike some LLM-generated text. <br> This Space is a small experiment to see if we can generate preference data for LLM-generated book blurbs. Whilst writing a blurb is very different from writing a whole book, it could be a neat experiment to see whether we can improve the ability of LLMs to generate book blurbs that people like! <br>"""
|
|
|
|
| 160 |
return f"Logged: {vote} by user {user_id}", gr.Row.update(visible=False)
|
| 161 |
|
| 162 |
|
| 163 |
+
short_description = """Vote on book blurbs generated by Large Language Models. Would you read the book the LLM generated based on the blurb? <br> Every five minutes, the dataset of votes created in this will be uploaded to the <a href="https://huggingface.co/datasets/davanstrien/summer-reading-preference">davanstrien/summer-reading-preference</a> dataset.
|
| 164 |
"""
|
| 165 |
|
| 166 |
full_description = """Large Language Models are already strong assistants for technical tasks like coding. \n\nIncreasingly, they are also being used to help with tasks like copywriting. The jury is out on whether the texts produced by language models in these applications are very appealing; "let\'s delve into" is a common example of the clunky cliche-ridden text that LLMs often produce. \n\n<br> However, there is growing interest in using LLMs to help with more creative tasks. Outside of larger companies, a growing community of people fine-tuning LLMs for all kinds of creative tasks. \n\nSome writers want to use LLMs – not as a replacement but as a companion – in their writing process. <br> One of the requirements for building models which are better able to generate responses people like is having preference data. Preference data come in many forms but essentially boil to a dataset that contains some kind of signal for whether people like or dislike some LLM-generated text. <br> This Space is a small experiment to see if we can generate preference data for LLM-generated book blurbs. Whilst writing a blurb is very different from writing a whole book, it could be a neat experiment to see whether we can improve the ability of LLMs to generate book blurbs that people like! <br>"""
|