vision benchmark
ODinW
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
Updated Aug 7, 2026
Published models16
Registry coverage16
MetricScore
EvidenceB
ODinW leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | AC | 51.8% | 100.0% | 16 | C | |
| 2 | AC | 51.1% | 93.3% | 16 | C | |
| 3 | AC | 50.8% | 86.7% | 16 | C | |
| 4 | AC | 48.6% | 80.0% | 16 | C | |
| 5 | AC | 48.2% | 73.3% | 16 | C | |
| 6 | AC | 47.5% | 66.7% | 16 | C | |
| 7 | AC | 46.6% | 60.0% | 16 | C | |
| 8 | AC | 44.7% | 53.3% | 16 | C | |
| 9 | AC | 44.5% | 46.7% | 16 | C | |
| 10 | AC | 43.2% | 40.0% | 16 | C | |
| 11 | AC | 42.6% | 33.3% | 16 | C | |
| 12 | AC | 42.4% | 26.7% | 16 | C | |
| 13 | AC | 42.3% | 20.0% | 16 | C | |
| 14 | AC | 41.1% | 13.3% | 16 | C | |
| 15 | AC | 39.8% | 6.7% | 16 | C | |
| 16 | AC | 39.4% | 0.0% | 16 | C |
Score distribution
Top published rows on the benchmark's original scale.
ODinW highlights
The top published results on this benchmark's own scale.
What is ODinW?
Definition and scoring fields from the benchmark registry.
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
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- ODinW
- Modality
- image
- Primary category
- vision
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- odinw|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about ODinW.
Which model scores highest on ODinW?
Qwen3.6 Plus is currently ranked first with 51.8%.
What does ODinW measure?
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
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
16 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.