Multimodal Evaluation on LLaVA-Bench
79.2LLaVA-Bench ScoreLLaVA-v1.6 (7B) w/ STIC
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LLaVA-v1.6 (7B) w/ STICBase Model=LLaVA-v1.6 (7B), Strategy=STIC2024.05 | 79.2 | — | |
| BaselineModel=Gemma 32026.05 | 79 | — | |
| LLaVA-v1.6 (7B)Model=LLaVA-v1.6 (7B)2024.05 | 77.3 | — | |
| SPHINX-2kResolution=2k2023.11 | 76.9 | — | |
| AsymVLMModel=Gemma 3, Text eviction ratio=90%2026.05 | 75.69 | — | |
| AsymVLMModel=Gemma 3, Text eviction ratio=75%2026.05 | 75.33 | — | |
| H2OModel=Gemma 3, Text eviction ratio=90%2026.05 | 75.3 | — | |
| AsymVLMModel=Gemma 3, Text eviction ratio=50%2026.05 | 74.69 | — | |
| H2OModel=Gemma 3, Text eviction ratio=75%2026.05 | 74.5 | — | |
| SPHINX-1kResolution=1k2023.11 | 74.3 | — | |
| SPHINX2023.11 | 73.5 | — | |
| RoboCodeXParameter size=13B2024.02 | 71.5 | — | |
| LLaVA-1.5 + TL-DPOBackbone=LLaVA-1.5-7B, Learning Objective=TL-DPO2025.06 | 71.2 | — | |
| LLaVA1.5-13BModel Size=13B2023.11 | 70.7 | — | |
| LLaVA-1.5Parameter size=13B2024.02 | 70.7 | — | |
| H2OModel=Gemma 3, Text eviction ratio=50%2026.05 | 69.83 | — | |
| LLaVA-v1.5 (7B) w/ STICBase Model=LLaVA-v1.5 (7B), Strategy=STIC2024.05 | 68.9 | — | |
| LLaVA-v1.5 (7B) w/ POVIDBase Model=LLaVA-v1.5 (7B), Strategy=POVID2024.05 | 68.7 | — | |
| LOVA3-7BTrain Paradigm=VQA, GenQA, EvalQA, LLM=Vicuna-7B2024.05 | 68.3 | — | |
| LOVA³-7BLLM=Vicuna-7B2024.05 | 67.7 | — | |
| LLaVA-1.5 + HALVABackbone=LLaVA-1.5-7B, Learning Objective=HALVA2025.06 | 67.2 | — | |
| LLaVA-1.5 + HA-DPOBackbone=LLaVA-1.5-7B, Learning Objective=HA-DPO2025.06 | 66.2 | — | |
| LLaVA-1.5 + RLHF-VBackbone=LLaVA-1.5-7B, Learning Objective=RLHF-V2025.06 | 65.4 | — | |
| LLaVA-v1.5 (7B)Model=LLaVA-v1.5 (7B)2024.05 | 65.4 | — | |
| LLaVA-1.5 + POVIDBackbone=LLaVA-1.5-7B, Learning Objective=POVID2025.06 | 64.8 | — | |
| LLaVA-1.5Train Paradigm=VQA, LLM=Vicuna-7B2024.05 | 64 | — | |
| LLaVA-1.5LLM=Vicuna-7B2024.05 | 64 | — | |
| LLaVA-1.5 + Human-PreferenceBackbone=LLaVA-1.5-7B, Learning Objective=Human-Preference2025.06 | 63.7 | — | |
| LLaVA-1.5Backbone=LLaVA-1.5-7B2025.06 | 63.4 | — | |
| LLaVA1.5-7BModel Size=7B2023.11 | 63.4 | — | |
| LLaVA-1.5Parameter size=7B2024.02 | 63.4 | — | |
| LLaVA-1.5 + VLfeedbackBackbone=LLaVA-1.5-7B, Learning Objective=VLfeedback2025.06 | 62.1 | — | |
| InstructBLIPParameter size=7B2024.02 | 60.9 | — | |
| InstructBLIPTrain Paradigm=VQA, VQG, LLM=Vicuna-7B2024.05 | 60.9 | — | |
| InstructBLIP (7B)Model=InstructBLIP (7B)2024.05 | 60.9 | — | |
| LOVA³-1.5BLLM=Phi-1.5B2024.05 | 60 | — | |
| mPLUG-OWL2 (7B)Model=mPLUG-OWL2 (7B)2024.05 | 59.9 | — | |
| LOVA3-1.5BTrain Paradigm=VQA, GenQA, EvalQA, LLM=Phi-1.5B2024.05 | 59.1 | — | |
| LLaVA-Phi-1.5Train Paradigm=VQA, LLM=Phi-1.5B2024.05 | 59 | — | |
| LLaVA-Phi-1.5LLM=Phi-1.5B2024.05 | 59 | — | |
| Shikra2023.11 | 58.8 | — | |
| StreamingLLMModel=Gemma 3, Text eviction ratio=90%2026.05 | 58.33 | — | |
| InstructBLIPParameter size=13B2024.02 | 58.2 | — | |
| InstructBLIPTrain Paradigm=VQA, VQG, LLM=Vicuna-13B2024.05 | 58.2 | — | |
| StreamingLLMModel=Gemma 3, Text eviction ratio=75%2026.05 | 56 | — | |
| StreamingLLMModel=Gemma 3, Text eviction ratio=50%2026.05 | 53 | — | |
| BLIP-22024.02 | 38.1 | — | |
| BLIP-2Train Paradigm=VQA, LLM=Vicuna-13B2024.05 | 38.1 | — | |
| DAMOModel=LLaVA1.5-7B2026.02 | — | 57.7 | |
| DAMOModel=Qwen2.5-VL-7B2026.02 | — | 87.4 | |
| DeCoModel=LLaVA1.5-7B2026.02 | — | 57 | |
| DeCoModel=Qwen2.5-VL-7B2026.02 | — | 86.8 | |
| DoLAModel=LLaVA1.5-7B2026.02 | — | 60.5 | |
| DoLAModel=Qwen2.5-VL-7B2026.02 | — | 66.2 | |
| EfficientLLaVAPruning Rate=0.32025.03 | — | 58.4 | |
| EfficientLLaVAPruning Rate=0.42025.03 | — | 54.5 | |
| EfficientLLaVAPruning Rate=0.52025.03 | — | 53 | |
| ICLAModel=LLaVA1.5-7B2026.02 | — | 61.9 | |
| ICLAModel=Qwen2.5-VL-7B2026.02 | — | 90.2 | |
| LLM-PrunerPruning Rate=0.32025.03 | — | 37.7 | |
| LLM-PrunerPruning Rate=0.42025.03 | — | 29.3 | |
| LLM-PrunerPruning Rate=0.52025.03 | — | 22.7 | |
| POVIDModel=LLaVA1.5-7B2026.02 | — | 60.2 | |
| SparseGPTPruning Rate=02025.03 | — | 55.9 | |
| SparseGPTPruning Rate=0.32025.03 | — | 57.9 | |
| SparseGPTPruning Rate=0.42025.03 | — | 52.5 | |
| SparseGPTPruning Rate=0.52025.03 | — | 48.8 | |
| VanillaModel=LLaVA1.5-7B2026.02 | — | 59.6 | |
| VanillaModel=Qwen2.5-VL-7B2026.02 | — | 87 | |
| VCDModel=LLaVA1.5-7B2026.02 | — | 60.6 | |
| VCDModel=Qwen2.5-VL-7B2026.02 | — | 88.7 | |
| VDDModel=LLaVA1.5-7B2026.02 | — | 59.4 | |
| VDDModel=Qwen2.5-VL-7B2026.02 | — | 87.2 |