|
|
import gradio as gr |
|
|
import pandas as pd |
|
|
|
|
|
|
|
|
data = { |
|
|
"Model": [ |
|
|
"DeepSeek-R1 (671B MoE)", "DeepSeek-V3 (671B MoE)", "Llama-3.1 (8B)", "Llama-3.1 (70B)", "Llama-3.1 (405B)", |
|
|
"Mistral (7B)", "Mixtral-8x22B (141B MoE)", "Qwen2.5 (7B)", "Qwen2.5 (72B)", |
|
|
"Claude 3.5 Haiku", "Claude 3.5 Sonnet", "Gemini 1.5 Flash", "Gemini 2.0 Flash", |
|
|
"Gemini 1.5 Pro", "Gemini 2.0 Pro", "GPT-4o", |
|
|
"LLaVa-v1.6-Mistral (7B)", |
|
|
"Gemini 1.5 Flash (MLLM)", "Gemini 2.0 Flash (MLLM)", "Gemini 1.5 Pro (MLLM)", "Gemini 2.0 Pro (MLLM)", "GPT-4o (MLLM)" |
|
|
], |
|
|
"T (Full)": [63.89, 64.91, 57.80, 63.37, 62.83, 49.94, 57.60, 56.06, 65.65, 56.94, 65.32, 56.90, 58.36, 62.78, 60.75, 57.43, None, 64.91, 64.79, 66.22, 66.42, 64.95], |
|
|
"TA (Full)": [45.81, 47.10, 40.69, 45.51, 45.82, 34.78, 39.86, 41.01, 47.42, 40.66, 43.38, 38.20, 37.47, 43.25, 44.59, 37.64, None, 45.06, 40.07, 46.90, 46.17, 44.53], |
|
|
"TAC (Full)": [21.29, 23.65, 19.08, 19.29, 22.67, 14.13, 16.49, 19.54, 19.65, 19.75, 22.54, 18.97, 19.20, 21.26, 18.63, 15.35, None, 20.66, 19.74, 23.23, 18.25, 19.60], |
|
|
"T (Segmented)": [71.18, 77.56, 71.86, 76.51, 78.60, 63.73, 70.12, 65.57, 79.20, 68.18, 75.90, 72.09, 70.05, 76.56, 74.66, 76.76, 16.46, 86.23, 82.24, 86.01, 86.04, 83.47], |
|
|
"TA (Segmented)": [47.29, 51.35, 43.42, 50.64, 49.50, 39.23, 41.75, 38.16, 48.36, 43.61, 46.47, 44.83, 42.11, 46.38, 47.18, 45.90, 11.45, 54.21, 48.00, 53.51, 54.28, 51.15], |
|
|
"TAC (Segmented)": [21.36, 28.17, 21.81, 23.96, 25.38, 19.83, 19.56, 20.76, 22.77, 21.54, 24.60, 21.31, 21.43, 22.87, 24.20, 24.50, 3.30, 23.27, 23.52, 24.97, 25.21, 27.86] |
|
|
} |
|
|
|
|
|
|
|
|
df = pd.DataFrame(data) |
|
|
|
|
|
def display_leaderboard(): |
|
|
return df |
|
|
|
|
|
|
|
|
with gr.Blocks(title="VideoConviction LLM Leaderboard") as demo: |
|
|
gr.Markdown(""" |
|
|
# VideoConviction: A Multimodal Benchmark for Human Conviction and Stock Market Recommendations |
|
|
|
|
|
**Paper**: [VideoConviction: A Multimodal Benchmark for Human Conviction and Stock Market Recommendations](https://doi.org/10.1145/3711896.3737417) |
|
|
**Conference**: ACM SIGKDD 2025 |
|
|
**Authors**: Michael Galarnyk, Veer Kejriwal, Agam Shah, Yash Bhardwaj, Nicholas Watney Meyer, Anand Krishnan, Sudheer Chava |
|
|
|
|
|
## Citation |
|
|
|
|
|
If you use this dataset, please cite our paper: |
|
|
|
|
|
```bibtex |
|
|
@inproceedings{galarnyk2025videoconviction, |
|
|
author = {Michael Galarnyk and Veer Kejriwal and Agam Shah and Yash Bhardwaj and Nicholas Watney Meyer and Anand Krishnan and Sudheer Chava}, |
|
|
title = {VideoConviction: A Multimodal Benchmark for Human Conviction and Stock Market Recommendations}, |
|
|
booktitle = {Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2 (KDD '25)}, |
|
|
year = {2025}, |
|
|
location = {Toronto, ON, Canada}, |
|
|
pages = {12}, |
|
|
publisher = {ACM}, |
|
|
doi = {10.1145/3711896.3737417} |
|
|
} |
|
|
``` |
|
|
|
|
|
# VideoConviction Benchmark Leaderboard |
|
|
This leaderboard shows the F1 scores of various LLMs and MLLMs across the VideoConviction benchmark tasks. |
|
|
""") |
|
|
gr.Dataframe(display_leaderboard, interactive=False) |
|
|
|
|
|
demo.launch() |
|
|
|
