Compositional Reasoning and Perception Evaluation on CRPE
94.15Exist ScoreIC-VCO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| IC-VCOContrastive Sample Source=Contrastive Sample Editing, Backbone=LLaVA-NeXT-Interleave-Qwen-7B2026.05 | 94.15 | 64.88 | |
| IC-VCOContrastive Sample Source=Synthetic, Backbone=LLaVA-NeXT-Interleave-Qwen-7B2026.05 | 93.16 | 65.63 | |
| S-VCOContrastive Sample Source=Contrastive Sample Editing, Backbone=LLaVA-NeXT-Interleave-Qwen-7B2026.05 | 92.58 | 65.89 | |
| SymMPOContrastive Sample Source=Contrastive Sample Editing, Backbone=LLaVA-NeXT-Interleave-Qwen-7B2026.05 | 92.47 | 65.58 | |
| mDPOContrastive Sample Source=Contrastive Sample Editing, Backbone=LLaVA-NeXT-Interleave-Qwen-7B2026.05 | 92.27 | 65.36 | |
| LLaVA-NeXT-Interleave-Qwen-7BBackbone=LLaVA-NeXT-Interleave-Qwen-7B2026.05 | 92.01 | 60.2 | |
| SymMPOContrastive Sample Source=Synthetic, Backbone=LLaVA-NeXT-Interleave-Qwen-7B2026.05 | 91.83 | 64.43 | |
| mDPOContrastive Sample Source=Synthetic, Backbone=LLaVA-NeXT-Interleave-Qwen-7B2026.05 | 91.79 | 64.84 | |
| S-VCOContrastive Sample Source=Synthetic, Backbone=LLaVA-NeXT-Interleave-Qwen-7B2026.05 | 89.92 | 66 | |
| DPOContrastive Sample Source=Synthetic, Backbone=LLaVA-NeXT-Interleave-Qwen-7B2026.05 | 89.55 | 65.38 | |
| V-DPOContrastive Sample Source=Synthetic, Backbone=LLaVA-NeXT-Interleave-Qwen-7B2026.05 | 89.41 | 65.22 | |
| DPOContrastive Sample Source=Contrastive Sample Editing, Backbone=LLaVA-NeXT-Interleave-Qwen-7B2026.05 | 89.35 | 63.31 | |
| V-DPOContrastive Sample Source=Contrastive Sample Editing, Backbone=LLaVA-NeXT-Interleave-Qwen-7B2026.05 | 89.2 | 62.88 |