Open-ended VQA on Personalization Evaluation Benchmark
0.716BLEU (Single)ICPT
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ICPTBackbone=LLaVA-NeXT-7B2026.05 | 0.716 | 0.697 | 0.702 | |
| PLVMBackbone=LLaVA-NeXT-7B2026.05 | 0.695 | 0.607 | 0.628 | |
| MC-LLaVABackbone=LLaVA-NeXT-7B2026.05 | 0.689 | 0.63 | 0.644 | |
| PeKitBackbone=LLaVA-NeXT-7B2026.05 | 0.68 | 0.524 | 0.561 | |
| GPT-4o + ICLBackbone=GPT-4o2026.05 | 0.663 | 0.605 | 0.619 | |
| RAPBackbone=LLaVA-NeXT-7B2026.05 | 0.651 | 0.529 | 0.558 | |
| PVITBackbone=LLaVA-NeXT-7B2026.05 | 0.642 | 0.537 | 0.562 | |
| Yo’ LLaVABackbone=LLaVA-NeXT-7B2026.05 | 0.593 | 0.451 | 0.485 | |
| ICLBackbone=LLaVA-NeXT-7B2026.05 | 0.572 | 0.483 | 0.504 | |
| MyVLMBackbone=LLaVA-NeXT-7B2026.05 | 0.548 | — | — | |
| VanillaBackbone=LLaVA-NeXT-7B2026.05 | 0.367 | 0.325 | 0.335 |