Finding top differences between tasks from the target model

Added explanation for the plot and a dataframe of the models

Added radar chart. Compares a model to the 5 models that have the closest performance on MMLU_average

added new results from hugging face

Added header back for the table

Added citation for the site

Changed streamlit to wide layout to see more of the table

Updated updated date

Added filter for parameter count. Fixed model filter so that it only filters on the Model name (index of the table)

Modified the selection of models and evaluations so that most do not show up by default. for a better user experience with 700+ models

Added search for Model name and Task name

Added reasoning for having scatter plots

Updated title now that there are over 700 open source models in the dataset

Updated with new results from hugging face

Added statement and hypothesis about moral scenarios

Plots have a default title

updated requirements.txt

Refactor of create_plot

Added finding from moral scenarios about threshold

Added bar chart for abstract algebra data.

updated hugging face results

Added introduction, links, and reduced the number of plots displayed

Updated with new data from hugging face

Modified download CSV feature so that the index column now has a title of model name

Added citation

Added new hugging face results

updated with new hugging face results

Add dashed line at the appropriate scale of the largest and smallest values on the plot so that plotly still zooms in to show that

Moved rank data into a separate method and dataframe

Added new huggingface results

Extracting parameter data from the names of the models

for the custom charts, remove columns with NaN values

Updated results

Refactor. Extracted methods.

Refactoring. Moved ResultDataProcessor class to a separate file to make it easier to use with experimentation in a jupyter notebook

Renamed class. Removed columns that were not useful.

Moved CSV download to just below the table

Added download CSV button

Added updated results from hugging face evaluation runs

Rename

Added custom scatterplot creation

Improving clarity. Moved MMLU average column to a more appropriate spot

Hiding filters unless box is selected. Removed model name column because it is the index of the table

Added a scatter plot with just the top 50 performing models on MMLU average

added MMLU overall average column. added a few charts comparing more moral reasoning and comparing MMLU overall to other data

Added statsmodels to be able to use a trendline in plotly

Updated data cleanup so that column names are cleaned up appropriatly with regex=True

added a few charts

fixed reversed plot. extracted making chart into a method

Update app.py and requirements.txt so that it will work with huggingface streamlit with the pandas 1.x version