Wikibench

ODinW

vision

Object Detection in the Wild (ODinW) benchmark for evaluating object detection models' task-level transfer ability across diverse real-world datasets in terms of prediction accuracy and adaptation efficiency

Leaderboard

Showing 15 of 15 results

  1. Qwen3.6 Plus
    51.8%
    i
  2. Qwen3.6-35B-A3B
    50.8%
    i
  3. Qwen3 VL 235B A22B Instruct
    48.6%
    i
  4. Qwen3 VL 4B Instruct
    48.2%
    i
  5. Qwen3 VL 30B A3B Instruct
    47.5%
    i
  6. Qwen3 VL 32B Instruct
    46.6%
    i
  7. Qwen3 VL 8B Instruct
    44.7%
    i
  8. Qwen3.5-122B-A10B
    44.5%
    i
  9. Qwen3 VL 235B A22B Thinking
    43.2%
    i
  10. Qwen3.5-35B-A3B
    42.6%
    i
  11. Qwen2.5-Omni-7B
    42.4%
    i
  12. Qwen3 VL 30B A3B Thinking
    42.3%
    i
  13. Qwen3.5-27B
    41.1%
    i
  14. Qwen3 VL 8B Thinking
    39.8%
    i
  15. Qwen3 VL 4B Thinking
    39.4%
    i