Visual Reasoning on HR-Bench (test)
69.94AccuracyVIRC-7B
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| VIRC-7B#Params=7B2025.12 | 69.94 | — | — | — | |
| VIRC-3B#Params=3B2025.12 | 66.75 | — | — | — | |
| Qwen2.5-VL-7B-Instruct#Params=7B2025.12 | 65.5 | — | — | — | |
| Qwen2.5-VL-3B-Instruct#Params=3B2025.12 | 65 | — | — | — | |
| Hint-GRPO-Qwen2-VL-7B#Params=7B2025.12 | 61.63 | — | — | — | |
| Hint-GRPO-Qwen2.5-VL-3B#Params=3B2025.12 | 60.94 | — | — | — | |
| GPT-4o2025.12 | 60.15 | — | — | — | |
| InternVL2.5-8B-MPO#Params=8B2025.12 | 59.69 | — | — | — | |
| InternVL2.5-8B#Params=8B2025.12 | 58.69 | — | — | — | |
| LLaVA-OV-Qwen2-7b-ov#Params=7B2025.12 | 57.6 | — | — | — | |
| DeepSeek-VL2#Params=4.5B2025.12 | 56.37 | — | — | — | |
| MM-Eureka#Params=7B2025.12 | 51.5 | — | — | — | |
| R1-VL-7B#Params=7B2025.12 | 50.38 | — | — | — | |
| MINT-CoT-7B#Params=7B2025.12 | 47.63 | — | — | — | |
| Qwen3-VL-8B-Instructw/ tool#Params=8B2025.12 | 44.87 | — | — | — | |
| DeepEyesSource=Reproduced through available model weights2025.12 | — | 75.5 | 71.3 | 73.4 | |
| DeepEyesv2Source=Reported in original or other relevant paper2025.12 | — | 77.9 | 73.8 | 75.9 | |
| DRIM2025.12 | — | 83.3 | 82.9 | 83.1 | |
| Gemini 2.5 Pro2025.12 | — | 68.5 | 60.1 | 64.3 | |
| GPT-52025.12 | — | 74.2 | 72.4 | 73.3 | |
| LLaVA-OneVisionSource=Reported in original or other relevant paper2025.12 | — | 61.2 | 54 | 57.6 | |
| Mini-o3Source=Reproduced through available model weights2025.12 | — | 73.5 | 73 | 73.3 | |
| Ovis2.5-9B2025.12 | — | 72.9 | 67.1 | 70 | |
| ThymeSource=Reported in original or other relevant paper2025.12 | — | 77 | 72 | 74.5 |