Uncertainty Quantification on NQ 16 additional samples (test)
85.6AUROCClustered Self-Assessment
Evaluation Results
| Method | Links | |
|---|---|---|
| Clustered Self-AssessmentBackbone=Qwen3-14B2026.06 | 85.6 | |
| Clustered Self-AssessmentBackbone=Qwen2.5-7B2026.06 | 85.1 | |
| Clustered Self-AssessmentBackbone=Qwen2.5-32B2026.06 | 85.1 | |
| Clustered Self-AssessmentBackbone=Gemma-3-12B2026.06 | 84.1 | |
| Clustered Self-AssessmentBackbone=Qwen3-30B-A3B2026.06 | 84 | |
| Clustered Self-AssessmentBackbone=Qwen2.5-14B2026.06 | 83.9 | |
| EigVBackbone=Qwen2.5-7B2026.06 | 83.4 | |
| Semantic EntropyBackbone=Qwen3-14B2026.06 | 82.9 | |
| EigVBackbone=Qwen3-14B2026.06 | 82.7 | |
| Semantic EntropyBackbone=Qwen2.5-7B2026.06 | 82 | |
| Clustered Self-AssessmentBackbone=Gemma-3-27B2026.06 | 82 | |
| EigVBackbone=Qwen2.5-32B2026.06 | 81.8 | |
| Semantic EntropyBackbone=Qwen2.5-32B2026.06 | 81.2 | |
| Semantic EntropyBackbone=Qwen3-30B-A3B2026.06 | 81 | |
| EigVBackbone=Qwen3-30B-A3B2026.06 | 81 | |
| DegBackbone=Gemma-3-12B2026.06 | 80.8 | |
| DegBackbone=Qwen2.5-7B2026.06 | 80.3 | |
| EigVBackbone=Qwen2.5-14B2026.06 | 80.2 | |
| P(True)Backbone=Qwen3-14B2026.06 | 79.9 | |
| Semantic EntropyBackbone=Qwen2.5-14B2026.06 | 79.6 | |
| DegBackbone=Qwen3-14B2026.06 | 79.6 | |
| SARBackbone=Qwen2.5-7B2026.06 | 79 | |
| SARBackbone=Qwen3-14B2026.06 | 79 | |
| DegBackbone=Gemma-3-27B2026.06 | 78.9 | |
| DegBackbone=Qwen2.5-32B2026.06 | 78.6 | |
| P(True)Backbone=Qwen2.5-32B2026.06 | 78.5 | |
| P(True)Backbone=Qwen3-30B-A3B2026.06 | 78.3 | |
| sentenceSARBackbone=Qwen2.5-7B2026.06 | 77.9 | |
| ProbabilityBackbone=Qwen2.5-7B2026.06 | 77.6 | |
| DegBackbone=Qwen3-30B-A3B2026.06 | 77.5 | |
| ProbabilityBackbone=Qwen3-14B2026.06 | 77.4 | |
| SARBackbone=Qwen2.5-32B2026.06 | 77.4 | |
| SARBackbone=Qwen3-30B-A3B2026.06 | 77.4 | |
| DegBackbone=Qwen2.5-14B2026.06 | 77.3 | |
| Predictive EntropyBackbone=Qwen2.5-7B2026.06 | 77 | |
| Eigen ScoreBackbone=Qwen2.5-7B2026.06 | 76.9 | |
| sentenceSARBackbone=Qwen3-14B2026.06 | 76.9 | |
| EccBackbone=Qwen2.5-7B2026.06 | 76.9 | |
| ProbabilityBackbone=Qwen2.5-32B2026.06 | 76.3 | |
| Predictive EntropyBackbone=Qwen3-14B2026.06 | 76.3 | |
| SARBackbone=Qwen2.5-14B2026.06 | 76.2 | |
| EccBackbone=Gemma-3-27B2026.06 | 76 | |
| sentenceSARBackbone=Qwen2.5-32B2026.06 | 75.9 | |
| EccBackbone=Qwen2.5-32B2026.06 | 75.9 | |
| EccBackbone=Qwen3-14B2026.06 | 75.9 | |
| LN EntropyBackbone=Qwen2.5-7B2026.06 | 75.7 | |
| Eigen ScoreBackbone=Qwen3-14B2026.06 | 75.6 | |
| EccBackbone=Gemma-3-12B2026.06 | 75.6 | |
| ProbabilityBackbone=Gemma-3-27B2026.06 | 75.5 | |
| tokenSARBackbone=Gemma-3-12B2026.06 | 75.4 | |
| LN EntropyBackbone=Qwen3-14B2026.06 | 75.2 | |
| Predictive EntropyBackbone=Qwen2.5-32B2026.06 | 75.1 | |
| sentenceSARBackbone=Qwen2.5-14B2026.06 | 75.1 | |
| EigVBackbone=Gemma-3-12B2026.06 | 75.1 | |
| Eigen ScoreBackbone=Qwen2.5-32B2026.06 | 75 | |
| tokenSARBackbone=Gemma-3-27B2026.06 | 75 | |
| PerplexityBackbone=Gemma-3-27B2026.06 | 74.9 | |
| sentenceSARBackbone=Qwen3-30B-A3B2026.06 | 74.9 | |
| PerplexityBackbone=Qwen2.5-7B2026.06 | 74.8 | |
| ProbabilityBackbone=Qwen3-30B-A3B2026.06 | 74.8 | |
| ProbabilityBackbone=Qwen2.5-14B2026.06 | 74.7 | |
| PerplexityBackbone=Gemma-3-12B2026.06 | 74.5 | |
| Predictive EntropyBackbone=Qwen2.5-14B2026.06 | 74.3 | |
| EccBackbone=Qwen2.5-14B2026.06 | 74.3 | |
| Predictive EntropyBackbone=Qwen3-30B-A3B2026.06 | 74.1 | |
| LN EntropyBackbone=Qwen3-30B-A3B2026.06 | 73.9 | |
| EccBackbone=Qwen3-30B-A3B2026.06 | 73.9 | |
| LN EntropyBackbone=Qwen2.5-32B2026.06 | 73.8 | |
| PerplexityBackbone=Qwen3-14B2026.06 | 73.7 | |
| Eigen ScoreBackbone=Qwen2.5-14B2026.06 | 73.5 | |
| EigVBackbone=Gemma-3-27B2026.06 | 73.5 | |
| Eigen ScoreBackbone=Qwen3-30B-A3B2026.06 | 73.3 | |
| tokenSARBackbone=Qwen2.5-7B2026.06 | 73.2 | |
| P(True)Backbone=Qwen2.5-14B2026.06 | 73.1 | |
| PerplexityBackbone=Qwen3-30B-A3B2026.06 | 72.9 | |
| ProbabilityBackbone=Gemma-3-12B2026.06 | 72.9 | |
| P(True)Backbone=Qwen2.5-7B2026.06 | 72.9 | |
| PerplexityBackbone=Qwen2.5-32B2026.06 | 72.8 | |
| Semantic EntropyBackbone=Gemma-3-12B2026.06 | 72.8 | |
| tokenSARBackbone=Qwen3-14B2026.06 | 72.5 | |
| LN EntropyBackbone=Qwen2.5-14B2026.06 | 72 | |
| tokenSARBackbone=Qwen2.5-32B2026.06 | 71.4 | |
| tokenSARBackbone=Qwen3-30B-A3B2026.06 | 71.4 | |
| SARBackbone=Gemma-3-12B2026.06 | 71 | |
| Semantic EntropyBackbone=Gemma-3-27B2026.06 | 71 | |
| PerplexityBackbone=Qwen2.5-14B2026.06 | 70.6 | |
| tokenSARBackbone=Qwen2.5-14B2026.06 | 70.1 | |
| sentenceSARBackbone=Gemma-3-12B2026.06 | 69.5 | |
| Predictive EntropyBackbone=Gemma-3-12B2026.06 | 68.8 | |
| sentenceSARBackbone=Gemma-3-27B2026.06 | 68.3 | |
| Predictive EntropyBackbone=Gemma-3-27B2026.06 | 68.1 | |
| SARBackbone=Gemma-3-27B2026.06 | 67.8 | |
| LN EntropyBackbone=Gemma-3-12B2026.06 | 67.3 | |
| Eigen ScoreBackbone=Gemma-3-12B2026.06 | 66.1 | |
| P(True)Backbone=Gemma-3-27B2026.06 | 65.9 | |
| Eigen ScoreBackbone=Gemma-3-27B2026.06 | 65.6 | |
| LN EntropyBackbone=Gemma-3-27B2026.06 | 64.9 | |
| P(True)Backbone=Gemma-3-12B2026.06 | 60.9 |