Affordance Reasoning on Affordance Dataset Setting B: S/U 1.0 (test)
85.4AccuracyGPT-5-mini
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-5-miniLatency (ms)=3,9002026.06 | 85.4 | |
| GPT-5.4Scale=medium, Latency (ms)=2,6462026.06 | 85.4 | |
| GPT-5.4Scale=low, Latency (ms)=1,8712026.06 | 83.9 | |
| GPT-5-nanoPrompting Strategy=standard, Latency (ms)=4,1002026.06 | 82.3 | |
| GPT-5-nanoPrompting Strategy=CoT, Latency (ms)=20,5002026.06 | 82.3 | |
| BLIPEvaluation Protocol=zero-shot, Latency (ms)=842026.06 | 70.8 | |
| A4DWeight Freezing Strategy=CLIP unfrozen, Latency (ms)=222026.06 | 60.8 | |
| A4DWeight Freezing Strategy=CLIP frozen-vision, Latency (ms)=222026.06 | 56.2 | |
| A4DWeight Freezing Strategy=CLIP frozen-text, Latency (ms)=222026.06 | 50.8 | |
| SigLIPEvaluation Protocol=zero-shot, Latency (ms)=202026.06 | 36.9 | |
| CLIPEvaluation Protocol=zero-shot, Latency (ms)=212026.06 | 30.8 | |
| ViLT-B32Evaluation Protocol=VQA, Latency (ms)=92026.06 | 18.2 |