Multimodal Understanding on Aggregate LLaVA 1.5 Suite
100Relative Average ScoreLLaVA-1.5-7B
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| LLaVA-1.5-7BBackbone=LLaVA-1.5-7B, Visual Token Budget=5762026.07 | 100 | — | — | — | |
| ViCA (3B)Base Model=LLaVA-1.5-3B2026.02 | 98.7 | — | — | — | |
| AnchorPruneBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.07 | 98.7 | — | — | — | |
| ViCA+PDrop† (3B)Base Model=LLaVA-1.5-3B2026.02 | 98.2 | — | — | — | |
| SparseVLMBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.07 | 97.3 | — | — | — | |
| VisionZipBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.07 | 97.3 | — | — | — | |
| ViCA (13B)Base Model=LLaVA-1.5-13B2026.02 | 97 | — | — | — | |
| DivPruneBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.07 | 96.5 | — | — | — | |
| ViCA+PDrop† (13B)Base Model=LLaVA-1.5-13B2026.02 | 96.3 | — | — | — | |
| AnchorPruneBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.07 | 96.2 | — | — | — | |
| CDPrunerBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.07 | 96.1 | — | — | — | |
| FastVBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.07 | 94.7 | — | — | — | |
| CDPrunerBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.07 | 94.2 | — | — | — | |
| AnchorPruneBackbone=LLaVA-1.5-7B, Visual Token Budget=322026.07 | 93.9 | — | — | — | |
| PyramidDropBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.07 | 93.4 | — | — | — | |
| PruMerge+Backbone=LLaVA-1.5-7B, Visual Token Budget=1282026.07 | 93.3 | — | — | — | |
| VisionZipBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.07 | 93 | — | — | — | |
| CDPrunerBackbone=LLaVA-1.5-7B, Visual Token Budget=322026.07 | 92.6 | — | — | — | |
| DivPruneBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.07 | 91.9 | — | — | — | |
| FastVBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.07 | 89.5 | — | — | — | |
| PruMerge+Backbone=LLaVA-1.5-7B, Visual Token Budget=642026.07 | 89.5 | — | — | — | |
| SparseVLMBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.07 | 89.1 | — | — | — | |
| DivPruneBackbone=LLaVA-1.5-7B, Visual Token Budget=322026.07 | 88.7 | — | — | — | |
| VisionZipBackbone=LLaVA-1.5-7B, Visual Token Budget=322026.07 | 86.1 | — | — | — | |
| PyramidDropBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.07 | 84.7 | — | — | — | |
| PruMerge+Backbone=LLaVA-1.5-7B, Visual Token Budget=322026.07 | 84.7 | — | — | — | |
| CoVFTBackbone=LLaVA-1.5-13B2026.03 | 62.9 | 69.66 | 65.11 | 56.17 | |
| FreezeBackbone=LLaVA-1.5-13B2026.03 | 61.43 | 68.78 | 64.27 | 53.84 | |
| BitFitBackbone=LLaVA-1.5-13B2026.03 | 61.43 | 68.55 | 64.23 | 54.05 | |
| Full fine-tuningBackbone=LLaVA-1.5-13B2026.03 | 61.3 | 68.66 | 64.43 | 53.54 | |
| CoVFTBackbone=LLaVA-1.5-7B2026.03 | 61.08 | 67.04 | 61.93 | 55.81 | |
| LLaVA-1.5Backbone=LLaVA-1.5-13B2026.03 | 60.73 | 67.79 | 64.62 | 52.75 | |
| BitFitBackbone=LLaVA-1.5-7B2026.03 | 59.57 | 66.14 | 61.58 | 53.1 | |
| Full fine-tuningBackbone=LLaVA-1.5-7B2026.03 | 59.29 | 66.69 | 61.29 | 52.17 | |
| LLaVA-1.5Backbone=LLaVA-1.5-7B2026.03 | 59.13 | 65.04 | 60.07 | 53.84 | |
| LoRABackbone=LLaVA-1.5-7B2026.03 | 59.04 | 65.93 | 60.86 | 52.45 | |
| FreezeBackbone=LLaVA-1.5-7B2026.03 | 58.93 | 66.23 | 61.2 | 51.71 | |
| SVPTBackbone=LLaVA-1.5-7B2026.03 | 58.7 | 64.83 | 61.08 | 52.37 | |
| VPTBackbone=LLaVA-1.5-7B2026.03 | 58.19 | 65.26 | 60.43 | 51.2 |