Real-world Question Answering on RealWorldQA
78.7Overall ScoreInternVL2.5-78B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| InternVL2.5-78BModel Name=InternVL2.5-78B2024.12 | 78.7 | — | |
| Qwen2-VL-72BModel Name=Qwen2-VL-72B2024.12 | 77.8 | — | |
| GPT-4o-20240513Model Name=GPT-4o-202405132024.12 | 75.4 | — | |
| InternVL2.5-26BModel Name=InternVL2.5-26B2024.12 | 74.5 | — | |
| InternVL2.5-38BModel Name=InternVL2.5-38B2024.12 | 73.5 | — | |
| InternVL2-Llama3-76BModel Name=InternVL2-Llama3-76B2024.12 | 72.2 | — | |
| LLaVA-OneVision-72BModel Name=LLaVA-OneVision-72B2024.12 | 71.9 | — | |
| InternVL2-40BModel Name=InternVL2-40B2024.12 | 71.8 | — | |
| Qwen2-VL-7BModel Name=Qwen2-VL-7B2024.12 | 70.1 | — | |
| InternVL2.5-8BModel Name=InternVL2.5-8B2024.12 | 70.1 | — | |
| InternVL2-26BModel Name=InternVL2-26B2024.12 | 68.3 | — | |
| Cambrian-34BModel Name=Cambrian-34B2024.12 | 67.8 | — | |
| Gemini-1.5-ProModel Name=Gemini-1.5-Pro2024.12 | 67.5 | — | |
| Ours-NonMathBudget=10%, Model=Qwen3-VL-2B2026.03 | 67.45 | — | |
| Ours-MixedBudget=10%, Model=Qwen3-VL-2B2026.03 | 67.32 | — | |
| InternVL-Chat-V1.5Model Name=InternVL-Chat-V1.52024.12 | 66 | — | |
| Ours-MathBudget=10%, Model=Qwen3-VL-2B2026.03 | 65.1 | — | |
| MiniCPM-V2.6Model Name=MiniCPM-V2.62024.12 | 65 | — | |
| InternVL2-8BModel Name=InternVL2-8B2024.12 | 64.4 | — | |
| InternVL2.5-4BModel Name=InternVL2.5-4B2024.12 | 64.3 | — | |
| Qwen2-VL-2BModel Name=Qwen2-VL-2B2024.12 | 62.6 | — | |
| InterlaceBudget=10%, Model=Qwen3-VL-2B2026.03 | 61.7 | — | |
| RandomBudget=10%, Model=Qwen3-VL-2B2026.03 | 61.44 | — | |
| GPT-4VModel Name=GPT-4V2024.12 | 61.4 | — | |
| InternVL2-4BModel Name=InternVL2-4B2024.12 | 60.7 | — | |
| InternVL2.5-2BModel Name=InternVL2.5-2B2024.12 | 60.1 | — | |
| Claude-3.5-SonnetModel Name=Claude-3.5-Sonnet2024.12 | 60.1 | — | |
| CKABudget=10%, Model=Qwen3-VL-2B2026.03 | 59.87 | — | |
| Ours-MathBudget=25%, Model=Qwen3-VL-2B2026.03 | 58.95 | — | |
| Ours-MixedBudget=25%, Model=Qwen3-VL-2B2026.03 | 58.95 | — | |
| Ours-NonMathBudget=25%, Model=Qwen3-VL-2B2026.03 | 58.43 | — | |
| CoVFTBackbone=LLaVA-1.5-13B2026.03 | 58.04 | — | |
| CoVFTBackbone=LLaVA-1.5-7B2026.03 | 57.52 | — | |
| InternVL2.5-1BModel Name=InternVL2.5-1B2024.12 | 57.5 | — | |
| InternVL2-2BModel Name=InternVL2-2B2024.12 | 57.3 | — | |
| BitFitBackbone=LLaVA-1.5-7B2026.03 | 57.12 | — | |
| SVPTBackbone=LLaVA-1.5-7B2026.03 | 56.86 | — | |
| FreezeBackbone=LLaVA-1.5-7B2026.03 | 56.73 | — | |
| LLaVA-1.5Backbone=LLaVA-1.5-7B2026.03 | 56.21 | — | |
| VPTBackbone=LLaVA-1.5-7B2026.03 | 56.21 | — | |
| FreezeBackbone=LLaVA-1.5-13B2026.03 | 55.94 | — | |
| Full fine-tuningBackbone=LLaVA-1.5-7B2026.03 | 55.82 | — | |
| LoRABackbone=LLaVA-1.5-7B2026.03 | 55.82 | — | |
| Full fine-tuningBackbone=LLaVA-1.5-13B2026.03 | 55.82 | — | |
| LLaVA-OneVision-0.5BModel Name=LLaVA-OneVision-0.5B2024.12 | 55.6 | — | |
| LLaVA-1.5Backbone=LLaVA-1.5-13B2026.03 | 55.29 | — | |
| RandomBudget=25%, Model=Qwen3-VL-2B2026.03 | 55.29 | — | |
| BitFitBackbone=LLaVA-1.5-13B2026.03 | 54.64 | — | |
| Phi-3.5-Vision-4BModel Name=Phi-3.5-Vision-4B2024.12 | 53.6 | — | |
| Ours-NonMathBudget=40%, Model=Qwen3-VL-2B2026.03 | 52.81 | — | |
| InternVL2-1BModel Name=InternVL2-1B2024.12 | 50.3 | — | |
| CKABudget=25%, Model=Qwen3-VL-2B2026.03 | 50.2 | — | |
| InterlaceBudget=40%, Model=Qwen3-VL-2B2026.03 | 48.24 | — | |
| InterlaceBudget=25%, Model=Qwen3-VL-2B2026.03 | 47.84 | — | |
| Ours-MixedBudget=40%, Model=Qwen3-VL-2B2026.03 | 47.19 | — | |
| Ours-MathBudget=40%, Model=Qwen3-VL-2B2026.03 | 45.88 | — | |
| CKABudget=40%, Model=Qwen3-VL-2B2026.03 | 40.26 | — | |
| RandomBudget=40%, Model=Qwen3-VL-2B2026.03 | 36.73 | — | |
| Claude-3.5-Sonnet2025.12 | — | 65.8 | |
| Claude-3.7-Sonnet2025.12 | — | 55.4 | |
| Claude-4-Sonnet2025.12 | — | 69.8 | |
| Gemini-1.5-Pro2025.12 | — | 67.5 | |
| Gemini-2.0-Flash2025.12 | — | 72.3 | |
| GPT-4.12025.12 | — | 78.7 | |
| GPT-4o2025.12 | — | 75.4 | |
| GPT-5-Mini2025.12 | — | 79 | |
| InternVL3-78B2025.12 | — | 78 | |
| InternVL3.5-38B2025.12 | — | 75.9 | |
| JARVISLLM=Vicuna-7B2025.12 | — | 55.6 | |
| JARVISLLM=Qwen2-7B2025.12 | — | 55.8 | |
| LLaVALLM=Vicuna-7B2025.12 | — | 54.9 | |
| LLaVALLM=Qwen2-7B2025.12 | — | 58 | |
| LLaVA-OneVision-72B2025.12 | — | 71.9 | |
| MastersBase Model=Qwen2.5-VL-7B2025.12 | — | 77.3 | |
| MastersBase Model=Qwen3-VL-8B2025.12 | — | 77.5 | |
| MastersBase Model=InternVL3-8B2025.12 | — | 74.8 | |
| MastersBase Model=InternVL3.5-8B2025.12 | — | 74.9 | |
| Molmo-72B2025.12 | — | 73.7 | |
| NVLM-72B2025.12 | — | 69.9 | |
| Ours w/o MaskingLLM=Vicuna-7B2025.12 | — | 55.7 | |
| Ours w/o MaskingLLM=Qwen2-7B2025.12 | — | 56.2 | |
| Qwen2.5-VL-72B2025.12 | — | 75.7 | |
| Qwen3-VL-32B2025.12 | — | 79 | |
| VIRALLLM=Vicuna-7B2025.12 | — | 54.8 | |
| VIRALLLM=Qwen2-7B2025.12 | — | 57.9 |