Information Visual Question Answering on InfoVQA
88.3AccuracyLlava-OV
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Llava-OVSize=7B2026.05 | 88.3 | — | |
| VL-Scaler-MiMOSize=7B2026.05 | 87.8 | — | |
| DeepEyesSize=7B2026.05 | 87.7 | — | |
| VL-ScalerSize=7B2026.05 | 87 | — | |
| Pixel ReasonerSize=7B2026.05 | 86.4 | — | |
| Qwen2.5-VL-InstructSize=72B2026.05 | 84.3 | — | |
| MiMO-VL-InstructSize=7B2026.05 | 84.2 | — | |
| Qwen2.5-VL-72B + CARESCost=-72%2025.10 | 84 | — | |
| BF16Model=Kimi-VL-A3B-Instruct, Avg-Bit=BF162026.06 | 83.61 | — | |
| GPT-4o-miniSize=-2026.05 | 83.3 | — | |
| BasePruning Ratio=100%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 83.21 | — | |
| ERASEPruning Ratio=-65.11%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 82.45 | — | |
| ERASEPruning Ratio=-72.77%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 82.45 | — | |
| IVC-PrunePruning Ratio=-65.11%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 82.21 | — | |
| ERASEPruning Ratio=-84.43%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 81.78 | — | |
| VanillaAvg. Tokens=100%2026.02 | 81.7 | — | |
| IVC-PrunePruning Ratio=-72.77%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 81.67 | — | |
| BF16Model=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=BF162026.06 | 81.46 | — | |
| IVC-PruneAvg. Tokens=50%2026.02 | 80.9 | — | |
| GPT-4oSize=-2026.05 | 80.7 | — | |
| Qwen2.5-VL-InstructSize=7B2026.05 | 80.7 | — | |
| BaseRatio=100%, Model Backbone=Qwen3-VL-4B2026.05 | 79.94 | 100 | |
| MODEModel=Kimi-VL-A3B-Instruct, Avg-Bit=W32026.06 | 79.87 | — | |
| VL-RethinkerSize=7B2026.05 | 79.5 | — | |
| BaselineModel=Llama Nemotron Nano 12B v2 VL2026.01 | 79.3 | — | |
| IVC-PruneRatio=-65.11%, Model Backbone=Qwen3-VL-4B2026.05 | 78.66 | 93.74 | |
| QADModel=Llama Nemotron Nano 12B v2 VL2026.01 | 78.4 | — | |
| Qwen3-VL-4B-InstructModel Category=Vision-Language Models2026.05 | 78.4 | — | |
| Pelican-UnifiedModel Category=Our Unified Model2026.05 | 78.4 | — | |
| QATModel=Llama Nemotron Nano 12B v2 VL2026.01 | 78.3 | — | |
| ERASERatio=-65.11%, Model Backbone=Qwen3-VL-4B2026.05 | 78.2 | 96.38 | |
| PTQModel=Llama Nemotron Nano 12B v2 VL2026.01 | 78.2 | — | |
| IVC-PruneRatio=-72.77%, Model Backbone=Qwen3-VL-4B2026.05 | 78 | 91.64 | |
| R1-VLSize=7B2026.05 | 78 | — | |
| ERASERatio=-72.77%, Model Backbone=Qwen3-VL-4B2026.05 | 77.57 | 94.09 | |
| ERASERatio=-84.43%, Model Backbone=Qwen3-VL-4B2026.05 | 77.43 | 89.39 | |
| OpenVLThinkerSize=7B2026.05 | 77.4 | — | |
| VEQ-MEModel=Kimi-VL-A3B-Instruct, Avg-Bit=W32026.06 | 77.24 | — | |
| IVC-PrunePruning Ratio=-84.43%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 76.67 | — | |
| MODEModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W32026.06 | 76.62 | — | |
| FastVAvg. Tokens=52%2026.02 | 76.4 | — | |
| mPLUG-Owl3Size=7B2026.05 | 76.3 | — | |
| BF16Precision=BF162026.06 | 76.27 | — | |
| MC-MoEModel=Kimi-VL-A3B-Instruct, Avg-Bit=W32026.06 | 75.53 | — | |
| GPT-4o2025.10 | 75 | — | |
| DocopilotSize=8B2026.05 | 75 | — | |
| HyperVL ViTLNumber of Parameters=2.0B2025.12 | 74.6 | — | |
| MoEQuantModel=Kimi-VL-A3B-Instruct, Avg-Bit=W32026.06 | 74.33 | — | |
| Claude-3.5Size=-2026.05 | 74.3 | — | |
| Speed-QModel=Kimi-VL-A3B-Instruct, Avg-Bit=W32026.06 | 74.27 | — | |
| AndesVLNumber of Parameters=2.4B2025.12 | 74.2 | — | |
| InternVL3-8B + CARESCost=-72%2025.10 | 74 | — | |
| InternVL3-8B + CARES-ARCost=-81%2025.10 | 74 | — | |
| MC-MoEModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W32026.06 | 73.65 | — | |
| SAIL-VL2Number of Parameters=2.7B2025.12 | 73.5 | — | |
| VEQ-MEModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W32026.06 | 73.23 | — | |
| Qwen2.5-VL-72B2025.10 | 73 | — | |
| GPT-4o + CARESCost=-76%2025.10 | 73 | — | |
| MBQModel=Kimi-VL-A3B-Instruct, Avg-Bit=W32026.06 | 72.69 | — | |
| Qwen3-VLNumber of Parameters=2.1B2025.12 | 72.4 | — | |
| InternVL3-8B2025.10 | 72 | — | |
| IVC-PruneRatio=-84.43%, Model Backbone=Qwen3-VL-4B2026.05 | 71.81 | 80.73 | |
| SAIL-VL1.5Number of Parameters=2.5B2025.12 | 71.8 | — | |
| MODEQuantization=W32026.06 | 71.53 | — | |
| Speed-QModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W32026.06 | 71.37 | — | |
| iLLaVAPruning Ratio=-65.11%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 71.09 | — | |
| MoEQuantModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W32026.06 | 70.81 | — | |
| InternVL3.5Number of Parameters=2.3B2025.12 | 70.8 | — | |
| HyperVLNumber of Parameters=1.8B2025.12 | 70.2 | — | |
| VEQ-MAQuantization=W32026.06 | 69.75 | — | |
| Ovis2Number of Parameters=2.5B2025.12 | 69.7 | — | |
| GPTQModel=Kimi-VL-A3B-Instruct, Avg-Bit=W32026.06 | 69.51 | — | |
| PDropAvg. Tokens=52%2026.02 | 69.4 | — | |
| MBQModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W32026.06 | 68.42 | — | |
| MODEModel=Kimi-VL-A3B-Instruct, Avg-Bit=W22026.06 | 67.75 | — | |
| MC-MoEQuantization=W32026.06 | 67.68 | — | |
| iLLaVAPruning Ratio=-72.77%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 67.14 | — | |
| PruneSIDPruning Ratio=-65.11%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 66.67 | — | |
| GPTQModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W32026.06 | 66.23 | — | |
| MODEModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W22026.06 | 66.19 | — | |
| InternVL3Number of Parameters=2.1B2025.12 | 66.1 | — | |
| Qwen2-VLNumber of Parameters=2.2B2025.12 | 65.5 | — | |
| GPTQQuantization=W32026.06 | 65.44 | — | |
| VEQ-MEModel=Kimi-VL-A3B-Instruct, Avg-Bit=W22026.06 | 64.94 | — | |
| PruneSIDRatio=-65.11%, Model Backbone=Qwen3-VL-4B2026.05 | 64.24 | 81.72 | |
| MC-MoEModel=Kimi-VL-A3B-Instruct, Avg-Bit=W22026.06 | 64.23 | — | |
| MODEQuantization=W22026.06 | 63.65 | — | |
| VEQ-MAQuantization=W22026.06 | 61.68 | — | |
| MM-R1-MGT-PerceReason2025.10 | 61.42 | — | |
| RECAPbackbone=Qwen2.5-VL-3B2025.10 | 60.78 | — | |
| iLLaVAPruning Ratio=-84.43%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 60.48 | — | |
| MC-MoEQuantization=W22026.06 | 60.1 | — | |
| Qwen2.5-VL-3Bvariant=MoDoMoDo2025.10 | 59.88 | — | |
| PruneSIDPruning Ratio=-72.77%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 59.69 | — | |
| DARTRatio=-65.11%, Model Backbone=Qwen3-VL-4B2026.05 | 59.34 | 84.8 | |
| VEQ-MEModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W22026.06 | 59.07 | — | |
| Vanilla RoPE2025.10 | 58.85 | — | |
| Qwen2.5-VL-3Bvariant=Uniform2025.10 | 58.3 | — | |
| vision-grpo-qwen-2.5-vl-3b2025.10 | 58.29 | — | |
| MRoPE-I2025.10 | 58.24 | — |