RAIVNLab/sugar-crepe
[NeurIPS 2023] A faithful benchmark for vision-language compositionality
PythonMIT
Issues
- 0
Negative caption generation prompts
#9 opened by samarth4149 - 1
- 0
Evaluation bug when using GELU vs QuickGELU -- changes the results for the trained models
#7 opened by bryant1410 - 0
Details about human performance
#5 opened by dubowsky - 4
- 5
Follow-up question to new NegCLIP variants
#2 opened by vishaal27 - 5
- 3
New NegCLIP variants
#1 opened by vishaal27