Cross-modal Retrieval on MIMIC-CXR
22.137Recall@1Entropy Regularization
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Entropy RegularizationVariant=Entropy, Setting=Random, N=100, Statistic=Mean ± SD2026.06 | 22.137 | 51.312 | 65.419 | 36.224 | |
| Entropy RegularizationVariant=Entropy, Setting=Random, N=100, Statistic=95% CI2026.06 | 21.823 | 50.882 | 65.001 | 35.898 | |
| head-only differentially private finetuningVariant=DP, Setting=Random, N=100, Statistic=Mean ± SD2026.06 | 9.416 | 29.398 | 43.859 | 20.501 | |
| head-only differentially private finetuningVariant=DP, Setting=Random, N=100, Statistic=95% CI2026.06 | 9.211 | 29.02 | 43.442 | 20.272 | |
| Entropy RegularizationVariant=Entropy, Setting=Random, N=1,000, Statistic=Mean ± SD2026.06 | 5.379 | 17.026 | 25.617 | 12.118 | |
| Entropy RegularizationVariant=Entropy, Setting=Random, N=1,000, Statistic=95% CI2026.06 | 5.232 | 16.705 | 25.213 | 11.923 | |
| head-only differentially private finetuningVariant=DP, Setting=Random, N=1,000, Statistic=Mean ± SD2026.06 | 1.547 | 6.153 | 10.413 | 4.937 | |
| head-only differentially private finetuningVariant=DP, Setting=Random, N=1,000, Statistic=95% CI2026.06 | 1.467 | 5.957 | 10.149 | 4.822 | |
| Entropy RegularizationVariant=Entropy, Setting=Random, N=10,000, Statistic=Mean ± SD2026.06 | 0.924 | 3.516 | 5.968 | 2.886 | |
| Entropy RegularizationVariant=Entropy, Setting=Random, N=10,000, Statistic=95% CI2026.06 | 0.855 | 3.372 | 5.78 | 2.8 | |
| Entropy RegularizationVariant=Entropy, Setting=Hard-neg, N=10,000, Statistic=Mean ± SD2026.06 | 0.682 | 2.67 | 4.727 | 2.313 | |
| Entropy RegularizationVariant=Entropy, Setting=Hard-neg, N=10,000, Statistic=95% CI2026.06 | 0.614 | 2.532 | 4.543 | 2.231 | |
| Entropy RegularizationVariant=Entropy, Setting=Random, N=43,793, Statistic=Mean ± SD2026.06 | 0.291 | 1.082 | 1.914 | 1.027 | |
| Entropy RegularizationVariant=Entropy, Setting=Random, N=43,793, Statistic=95% CI2026.06 | 0.24 | 0.989 | 1.79 | 0.971 | |
| head-only differentially private finetuningVariant=DP, Setting=Random, N=10,000, Statistic=Mean ± SD2026.06 | 0.19 | 0.874 | 1.662 | 0.895 | |
| head-only differentially private finetuningVariant=DP, Setting=Random, N=10,000, Statistic=95% CI2026.06 | 0.161 | 0.799 | 1.559 | 0.851 | |
| head-only differentially private finetuningVariant=DP, Setting=Hard-neg, N=10,000, Statistic=Mean ± SD2026.06 | 0.158 | 0.718 | 1.416 | 0.785 | |
| head-only differentially private finetuningVariant=DP, Setting=Hard-neg, N=10,000, Statistic=95% CI2026.06 | 0.125 | 0.641 | 1.312 | 0.741 | |
| head-only differentially private finetuningVariant=DP, Setting=Random, N=43,786, Statistic=Mean ± SD2026.06 | 0.039 | 0.216 | 0.418 | 0.264 | |
| head-only differentially private finetuningVariant=DP, Setting=Random, N=43,786, Statistic=95% CI2026.06 | 0.021 | 0.174 | 0.361 | 0.242 |