Agentic Reasoning on VisualProbe Medium
84.7AccuracyDeepEyes-7B
Evaluation results
| Method | Links | |
|---|---|---|
| DeepEyes-7BActivation Replay=true2025.11 | 84.7 | |
| DeepEyes-7BActivation Replay=false2025.11 | 82.1 | |
| Qwen2.5-VL-7BActivation Replay=false2025.11 | 72 |
| Method | Links | |
|---|---|---|
| DeepEyes-7BActivation Replay=true2025.11 | 84.7 | |
| DeepEyes-7BActivation Replay=false2025.11 | 82.1 | |
| Qwen2.5-VL-7BActivation Replay=false2025.11 | 72 |