Scientific Reasoning on ScienceQA
96.21ScoreVISE
Evaluation results
| Method | Links | |
|---|---|---|
| VISEModel Scale=32B2026.06 | 96.21 | |
| VisionZero-ChartModel Scale=32B2026.06 | 96.11 | |
| InternVL3.5parameters=8B2026.01 | 96.09 | |
| VisionZero-CLEVRModel Scale=32B2026.06 | 96.05 | |
| Intern-S1variant=mini, parameters=9B2026.01 | 96.04 | |
| iReasonerModel Scale=32B2026.06 | 96.03 | |
| EvoLMMModel Scale=32B2026.06 | 95.98 | |
| VisionZero-RWModel Scale=32B2026.06 | 95.84 | |
| VisPlayModel Scale=32B2026.06 | 95.8 | |
| BaseModel Scale=32B2026.06 | 95.68 | |
| LLaVA-OneVisionLLM=Qwen2-7B, Tokens=–, Zero-shot=true2025.07 | 95.4 | |
| MiniCPM-Vversion=4.5, parameters=8B2026.01 | 95.07 | |
| LLaVA-OVversion=1.5, parameters=8B2026.01 | 94.86 | |
| InternVL3.5-REGATELLM=Qwen3-14B, Tokens=2.32B (↓ 41.41%), Zero-shot=true2025.07 | 94.4 | |
| InternVL3.5LLM=Qwen3-14B, Tokens=3.96B, Zero-shot=true2025.07 | 93.3 | |
| Qwen3-VLparameters=8B2026.01 | 92.95 | |
| VISEModel Scale=8B2026.06 | 92.81 | |
| GPT-4.1LLM=–, Tokens=–, Zero-shot=true2025.07 | 92.8 | |
| VisionZero-ChartModel Scale=8B2026.06 | 92.41 | |
| VisionZero-CLEVRModel Scale=8B2026.06 | 92.35 | |
| iReasonerModel Scale=8B2026.06 | 92.11 | |
| EvoLMMModel Scale=8B2026.06 | 91.96 | |
| Innovator-VLvariant=8B-Thinking, parameters=8B2026.01 | 91.65 | |
| Innovator-VLvariant=8B-Instruct, parameters=8B2026.01 | 91.49 | |
| VisionZero-RWModel Scale=8B2026.06 | 91.18 | |
| VisPlayModel Scale=8B2026.06 | 91.02 | |
| Claude3.7-SonnetLLM=–, Tokens=–, Zero-shot=true2025.07 | 90.9 | |
| BaseModel Scale=8B2026.06 | 90.88 | |
| GPT-4oLLM=–, Tokens=–, Zero-shot=true2025.07 | 90.1 | |
| VISEModel Scale=4B2026.06 | 90.04 | |
| MiMo-VLtraining=7B-RL, parameters=7B2026.01 | 89.15 | |
| Qwen2.5-VLLLM=Qwen2.5-7B, Tokens=–, Zero-shot=true2025.07 | 89 | |
| MiMo-VLtraining=7B-SFT, parameters=7B2026.01 | 88.89 | |
| iReasonerModel Scale=4B2026.06 | 88.62 | |
| VisionZero-ChartModel Scale=4B2026.06 | 88.5 | |
| VisionZero-CLEVRModel Scale=4B2026.06 | 88.46 | |
| EvoLMMModel Scale=4B2026.06 | 88.46 | |
| VisionZero-RWModel Scale=4B2026.06 | 87.73 | |
| BaseModel Scale=4B2026.06 | 87.51 | |
| VisPlayModel Scale=4B2026.06 | 87.3 | |
| Gemini-1.5-ProLLM=–, Tokens=–, Zero-shot=true2025.07 | 85.7 | |
| VISEModel Scale=2B2026.06 | 83.61 | |
| IADA +LoRATrainable Params=7.7M2026.03 | 83.4 | |
| Gemini-1.5-FlashLLM=–, Tokens=–, Zero-shot=true2025.07 | 83.3 | |
| iReasonerModel Scale=2B2026.06 | 83.12 | |
| EvoLMMModel Scale=2B2026.06 | 83.01 | |
| Pre-trainedTrainable Params=02026.03 | 82.9 | |
| AttnRes+LoRATrainable Params=7.7M2026.03 | 82.6 | |
| VisionZero-CLEVRModel Scale=2B2026.06 | 81.96 | |
| VisPlayModel Scale=2B2026.06 | 80.74 | |
| VideoLLaMA2-REGATELLM=Qwen2-7B, Tokens=49.27M (↓ 41.22%), Zero-shot=true2025.07 | 80.5 | |
| VisionZero-ChartModel Scale=2B2026.06 | 79.93 | |
| VisionZero-RWModel Scale=2B2026.06 | 79.9 | |
| BaseModel Scale=2B2026.06 | 79.42 | |
| VILA1.5LLM=Llama-2-13B, Tokens=–, Zero-shot=true2025.07 | 79.1 | |
| LLaVA-NextLLM=Mistral-7B, Tokens=–, Zero-shot=true2025.07 | 73 | |
| LoRA-onlyTrainable Params=7.6M2026.03 | 72.1 | |
| LLaVA-1.6LLM=Vicuna-7B, Tokens=–, Zero-shot=true2025.07 | 70.1 | |
| VanillaBackbone=LLaVA-1.5-7B, Visual Token Budget=5762026.06 | 69.5 | |
| V2DropBackbone=LLaVA-1.5-7B, Visual Token Budget=1922026.06 | 69.3 | |
| PDropBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.06 | 69.2 | |
| SparseVLMBackbone=LLaVA-1.5-7B, Visual Token Budget=1922026.06 | 69.1 | |
| PDropBackbone=LLaVA-1.5-7B, Visual Token Budget=1922026.06 | 69 | |
| VisionZipBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.06 | 69 | |
| VisionZipBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.06 | 68.9 | |
| V2DropBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.06 | 68.9 | |
| ApETBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.06 | 68.9 | |
| V2DropBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.06 | 68.8 | |
| EvoCutBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.06 | 68.8 | |
| VisionZipBackbone=LLaVA-1.5-7B, Visual Token Budget=1922026.06 | 68.7 | |
| ApETBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.06 | 68.7 | |
| EvoCutBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.06 | 68.7 | |
| ApETBackbone=LLaVA-1.5-7B, Visual Token Budget=1922026.06 | 68.5 | |
| EvoCutBackbone=LLaVA-1.5-7B, Visual Token Budget=1922026.06 | 68.5 | |
| PDropBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.06 | 68.4 | |
| Qwen-VL-ChatLLM=Qwen-7B, Tokens=–, Zero-shot=true2025.07 | 68.2 | |
| FastVBackbone=LLaVA-1.5-7B, Visual Token Budget=1922026.06 | 67.3 | |
| SparseVLMBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.06 | 67.1 | |
| LLaVA-1.5LLM=Vicuna-7B, Tokens=–, Zero-shot=true2025.07 | 66.8 | |
| SparseVLMBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.06 | 62.2 | |
| VideoLLaMA2LLM=Qwen2-7B, Tokens=83.82M, Zero-shot=true2025.07 | 61.4 | |
| InstructBLIPLLM=Vicuna-7B, Tokens=–, Zero-shot=true2025.07 | 60.5 | |
| FastVBackbone=LLaVA-1.5-7B, Visual Token Budget=1282026.06 | 60.2 | |
| FastVBackbone=LLaVA-1.5-7B, Visual Token Budget=642026.06 | 51.1 | |
| VideoChat2-REGATELLM=Mistral-7B, Tokens=2.22B (↓ 43.51%), Zero-shot=true2025.07 | 46.6 | |
| VideoChat2LLM=Mistral-7B, Tokens=3.93B, Zero-shot=true2025.07 | 40.8 |