Spaces:
Running on Zero
Running on Zero
Update app.py
Browse files
app.py
CHANGED
|
@@ -80,7 +80,7 @@ with gr.Blocks() as iface:
|
|
| 80 |
# 📈 Predict Academic Impact of Newly Published Paper!
|
| 81 |
### Estimate the future academic impact from the title and abstract with LLM.
|
| 82 |
###### [Full Paper](https://arxiv.org/abs/2408.03934)
|
| 83 |
-
######
|
| 84 |
""")
|
| 85 |
with gr.Row():
|
| 86 |
with gr.Column():
|
|
@@ -101,8 +101,8 @@ with gr.Blocks() as iface:
|
|
| 101 |
gr.Markdown("""
|
| 102 |
## Ethical Warnings and Important Notes
|
| 103 |
- It is intended as a tool **for research and educational purposes only**.
|
| 104 |
-
- Please refrain from deliberately embellishing the title and abstract to boost scores, and avoid making false claims.
|
| 105 |
-
- Our training data only includes samples from the fields including cs.CV, cs.CL (NLP), and cs.AI. Predictions outside these areas are not recommended for reference.
|
| 106 |
- The **predicted value** is a probability generated by the model and **does NOT reflect paper quality or novelty**.
|
| 107 |
- To identify potentially impactful papers, this study uses the sigmoid+MSE approach to optimize NDCG values (over sigmoid+BCE), resulting in predicted values generally concentrated **between 0.1 and 0.9**.
|
| 108 |
- Empirically, it is considered a predicted influence score greater than **0.65** to indicate an impactful paper.
|
|
|
|
| 80 |
# 📈 Predict Academic Impact of Newly Published Paper!
|
| 81 |
### Estimate the future academic impact from the title and abstract with LLM.
|
| 82 |
###### [Full Paper](https://arxiv.org/abs/2408.03934)
|
| 83 |
+
###### Please be advised: Local inference of the proposed method is instant, but ZeroGPU requires quantized model reinitialization with each "Predict", causing slight delays. (typically wont take more than 30 secs)
|
| 84 |
""")
|
| 85 |
with gr.Row():
|
| 86 |
with gr.Column():
|
|
|
|
| 101 |
gr.Markdown("""
|
| 102 |
## Ethical Warnings and Important Notes
|
| 103 |
- It is intended as a tool **for research and educational purposes only**.
|
| 104 |
+
- Please refrain from deliberately embellishing the title and abstract to boost scores, and **avoid making false claims**.
|
| 105 |
+
- Our **training data only includes** samples from the fields including **cs.CV, cs.CL (NLP), and cs.AI**. Predictions outside these areas are not recommended for reference.
|
| 106 |
- The **predicted value** is a probability generated by the model and **does NOT reflect paper quality or novelty**.
|
| 107 |
- To identify potentially impactful papers, this study uses the sigmoid+MSE approach to optimize NDCG values (over sigmoid+BCE), resulting in predicted values generally concentrated **between 0.1 and 0.9**.
|
| 108 |
- Empirically, it is considered a predicted influence score greater than **0.65** to indicate an impactful paper.
|