Uncertainty Estimation on Landmarks
69.31AUROCLayer Top-m Prob.
Evaluation Results
| Method | Links | |
|---|---|---|
| Layer Top-m Prob.Probe Type=Pretrained, Probe Architecture=Linear, Feature Representation=Layer Top-m Prob.2026.06 | 69.31 | |
| Ensemble (task)Probe Type=Pretrained, Probe Architecture=Linear, Feature Representation=Ensemble of probes on pooled data2026.06 | 69.25 | |
| Internal VarianceProbe Type=Pretrained, Probe Architecture=Linear, Feature Representation=Internal Variance2026.06 | 67.74 | |
| Ensemble (probe)Probe Type=Pretrained, Probe Architecture=Linear, Feature Representation=Ensemble of three Internal variance probes2026.06 | 64.46 | |
| Lookback RatioProbe Type=Pretrained, Probe Architecture=Linear, Feature Representation=Lookback Ratio2026.06 | 62.72 | |
| Linear (#256)Probe Type=Baseline, Probe Architecture=Linear, Number of Training Samples=256, Feature Representation=Embedding (last)2026.06 | 61.28 | |
| Linear (#128)Probe Type=Baseline, Probe Architecture=Linear, Number of Training Samples=128, Feature Representation=Embedding (last)2026.06 | 61.24 | |
| Embedding (last)Probe Type=Pretrained, Probe Architecture=Linear, Feature Representation=Embedding (last)2026.06 | 61.01 | |
| Linear (#64)Probe Type=Baseline, Probe Architecture=Linear, Number of Training Samples=64, Feature Representation=Embedding (last)2026.06 | 58.27 | |
| MLP (#128)Probe Type=Baseline, Probe Architecture=MLP, Number of Training Samples=128, Feature Representation=Embedding (last)2026.06 | 57.51 |