Image-Text Matching on Winoground (full)
52.7AccuracyVisReP
Evaluation Results
| Method | Links | |
|---|---|---|
| VisRePLLM=CodeLlama-7B, Backbone=ViperGPT [29]2024.04 | 52.7 | |
| Frozen Proprietary LLMLLM=GPT-3.5-turbo, Backbone=ViperGPT [29]2024.04 | 45.6 | |
| Frozen Open LLMLLM=CodeLlama-7B, Backbone=ViperGPT [29]2024.04 | 41.3 |