Interleaved Image-Text Composition User Study 20 articles 1.0 (test)
77Human Avg ScoreGPT4-V
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| GPT4-VImage Selection=Context-based selection2023.09 | 77 | 73 | 95 | 64 | 89 | 88 | 96 | 78 | |
| InternLM-XComposerImage Selection=Context-based selection2023.09 | 74 | 65 | 90 | 67 | 87 | 90 | 93 | 79 | |
| GPT3.5Image Selection=Top-1 Retrieval2023.09 | 71 | 64 | 88 | 59 | 84 | 84 | 91 | 75 | |
| Intern-20BImage Selection=Top-1 Retrieval2023.09 | 64 | 54 | 82 | 56 | 81 | 82 | 89 | 70 | |
| Qwen-14BImage Selection=Top-1 Retrieval2023.09 | 62 | 52 | 80 | 54 | 82 | 78 | 91 | 72 |