DanielePoterti commited on
Commit
4ad3a38
1 Parent(s): ac2185f

Upload 2 files

Browse files
Files changed (2) hide show
  1. src/macro_area.csv +1 -0
  2. src/question_format.csv +1 -0
src/macro_area.csv CHANGED
@@ -14,6 +14,7 @@ dolphin-llama-3-70b,82.4,84.9,78.8,85.7,55.2,50.0,0.0,68.4,83.3
14
  gemini-flash-1.5,83.3,85.5,81.8,85.7,62.1,83.3,25.0,63.2,66.7
15
  gemini-pro,78.7,82.1,81.8,71.4,51.7,70.8,0.0,68.4,66.7
16
  gemini-pro-1.5,90.7,87.7,84.8,57.1,55.2,58.3,25.0,63.2,33.3
 
17
  gemma-2-9b-it,75.9,82.7,66.7,71.4,51.7,58.3,0.0,57.9,83.3
18
  gpt-3.5-turbo-0125,61.1,64.8,63.6,42.9,55.2,58.3,0.0,47.4,83.3
19
  gpt-4-turbo,86.1,89.9,81.8,71.4,86.2,79.2,50.0,73.7,100.0
 
14
  gemini-flash-1.5,83.3,85.5,81.8,85.7,62.1,83.3,25.0,63.2,66.7
15
  gemini-pro,78.7,82.1,81.8,71.4,51.7,70.8,0.0,68.4,66.7
16
  gemini-pro-1.5,90.7,87.7,84.8,57.1,55.2,58.3,25.0,63.2,33.3
17
+ gemma-2-27b-it,81.5,88.8,78.8,85.7,62.1,62.5,0.0,73.7,66.7
18
  gemma-2-9b-it,75.9,82.7,66.7,71.4,51.7,58.3,0.0,57.9,83.3
19
  gpt-3.5-turbo-0125,61.1,64.8,63.6,42.9,55.2,58.3,0.0,47.4,83.3
20
  gpt-4-turbo,86.1,89.9,81.8,71.4,86.2,79.2,50.0,73.7,100.0
src/question_format.csv CHANGED
@@ -14,6 +14,7 @@ dolphin-llama-3-70b,100.0,50.0,100.0,91.7,28.6,0.0,50.0,93.0,71.4,44.4,81.5,0.0,
14
  gemini-flash-1.5,90.6,0.0,0.0,86.7,71.4,100.0,33.3,93.0,85.7,88.9,88.9,0.0,100.0,50.0,81.2,38.5,50.0,81.0,0.0
15
  gemini-pro,96.9,0.0,0.0,90.0,14.3,0.0,16.7,80.3,71.4,66.7,88.9,0.0,100.0,0.0,79.2,46.2,64.3,69.0,0.0
16
  gemini-pro-1.5,96.9,0.0,0.0,90.0,42.9,100.0,33.3,87.3,42.9,77.8,87.6,0.0,100.0,50.0,79.2,46.2,85.7,85.7,12.5
 
17
  gemma-2-9b-it,96.9,50.0,0.0,88.3,14.3,0.0,33.3,83.1,42.9,66.7,77.8,0.0,50.0,0.0,79.2,61.5,57.1,61.9,12.5
18
  gpt-3.5-turbo-0125,84.4,0.0,0.0,73.3,14.3,0.0,50.0,53.5,42.9,44.4,67.9,0.0,75.0,50.0,68.8,46.2,71.4,52.4,0.0
19
  gpt-4-turbo,100.0,100.0,100.0,91.7,71.4,100.0,66.7,87.3,100.0,100.0,92.6,100.0,100.0,50.0,89.6,53.8,57.1,81.0,12.5
 
14
  gemini-flash-1.5,90.6,0.0,0.0,86.7,71.4,100.0,33.3,93.0,85.7,88.9,88.9,0.0,100.0,50.0,81.2,38.5,50.0,81.0,0.0
15
  gemini-pro,96.9,0.0,0.0,90.0,14.3,0.0,16.7,80.3,71.4,66.7,88.9,0.0,100.0,0.0,79.2,46.2,64.3,69.0,0.0
16
  gemini-pro-1.5,96.9,0.0,0.0,90.0,42.9,100.0,33.3,87.3,42.9,77.8,87.6,0.0,100.0,50.0,79.2,46.2,85.7,85.7,12.5
17
+ gemma-2-27b-it,96.9,50.0,0.0,90.0,71.4,0.0,33.3,88.7,71.4,77.8,82.7,0.0,100.0,0.0,85.4,69.2,64.3,71.4,25.0
18
  gemma-2-9b-it,96.9,50.0,0.0,88.3,14.3,0.0,33.3,83.1,42.9,66.7,77.8,0.0,50.0,0.0,79.2,61.5,57.1,61.9,12.5
19
  gpt-3.5-turbo-0125,84.4,0.0,0.0,73.3,14.3,0.0,50.0,53.5,42.9,44.4,67.9,0.0,75.0,50.0,68.8,46.2,71.4,52.4,0.0
20
  gpt-4-turbo,100.0,100.0,100.0,91.7,71.4,100.0,66.7,87.3,100.0,100.0,92.6,100.0,100.0,50.0,89.6,53.8,57.1,81.0,12.5