Uncertainty Quantification on TQA
93.3AUROCClustered Self-Assessment
Evaluation Results
| Method | Links | |
|---|---|---|
| Clustered Self-AssessmentModel=Qwen2.5-32B, Number of additional samples=22026.06 | 93.3 | |
| Clustered Self-AssessmentModel=Qwen2.5-14B, Number of additional samples=22026.06 | 93.2 | |
| Clustered Self-AssessmentModel=Qwen3-14B, Number of additional samples=22026.06 | 92.7 | |
| Clustered Self-AssessmentModel=Qwen3-30B-A3B, Number of additional samples=22026.06 | 92.3 | |
| Clustered Self-AssessmentModel=Gemma-3-12B, Number of additional samples=22026.06 | 92.2 | |
| Clustered Self-AssessmentModel=Gemma-3-27B, Number of additional samples=22026.06 | 91.9 | |
| Clustered Self-AssessmentModel=Qwen2.5-7B, Number of additional samples=22026.06 | 91.5 | |
| SARModel=Gemma-3-12B, Number of additional samples=22026.06 | 91.4 | |
| SARModel=Gemma-3-27B, Number of additional samples=22026.06 | 90.1 | |
| ProbabilityModel=Qwen3-14B, Number of additional samples=22026.06 | 89.7 | |
| tokenSARModel=Gemma-3-12B, Number of additional samples=22026.06 | 89.7 | |
| PerplexityModel=Qwen3-14B, Number of additional samples=22026.06 | 89.5 | |
| PerplexityModel=Gemma-3-12B, Number of additional samples=22026.06 | 89.5 | |
| LN EntropyModel=Gemma-3-12B, Number of additional samples=22026.06 | 89.5 | |
| sentenceSARModel=Qwen3-14B, Number of additional samples=22026.06 | 89.5 | |
| PerplexityModel=Gemma-3-27B, Number of additional samples=22026.06 | 89.4 | |
| Predictive EntropyModel=Qwen3-14B, Number of additional samples=22026.06 | 89.3 | |
| LN EntropyModel=Qwen3-14B, Number of additional samples=22026.06 | 89.3 | |
| DegModel=Gemma-3-12B, Number of additional samples=22026.06 | 89.3 | |
| tokenSARModel=Gemma-3-27B, Number of additional samples=22026.06 | 89.1 | |
| ProbabilityModel=Qwen2.5-7B, Number of additional samples=22026.06 | 89 | |
| P(True)Model=Qwen2.5-32B, Number of additional samples=22026.06 | 89 | |
| PerplexityModel=Qwen2.5-7B, Number of additional samples=22026.06 | 88.8 | |
| EigVModel=Gemma-3-12B, Number of additional samples=22026.06 | 88.8 | |
| sentenceSARModel=Gemma-3-27B, Number of additional samples=22026.06 | 88.7 | |
| P(True)Model=Qwen3-14B, Number of additional samples=22026.06 | 88.6 | |
| sentenceSARModel=Gemma-3-12B, Number of additional samples=22026.06 | 88.6 | |
| sentenceSARModel=Qwen2.5-7B, Number of additional samples=22026.06 | 88.5 | |
| LN EntropyModel=Qwen2.5-7B, Number of additional samples=22026.06 | 88.4 | |
| SARModel=Qwen3-14B, Number of additional samples=22026.06 | 88.4 | |
| ProbabilityModel=Qwen2.5-32B, Number of additional samples=22026.06 | 88.3 | |
| ProbabilityModel=Gemma-3-27B, Number of additional samples=22026.06 | 88.3 | |
| Predictive EntropyModel=Qwen2.5-7B, Number of additional samples=22026.06 | 88.3 | |
| ProbabilityModel=Qwen2.5-14B, Number of additional samples=22026.06 | 88.2 | |
| LN EntropyModel=Gemma-3-27B, Number of additional samples=22026.06 | 88.1 | |
| sentenceSARModel=Qwen2.5-32B, Number of additional samples=22026.06 | 88.1 | |
| PerplexityModel=Qwen2.5-14B, Number of additional samples=22026.06 | 88 | |
| sentenceSARModel=Qwen2.5-14B, Number of additional samples=22026.06 | 88 | |
| PerplexityModel=Qwen2.5-32B, Number of additional samples=22026.06 | 87.9 | |
| Predictive EntropyModel=Qwen2.5-32B, Number of additional samples=22026.06 | 87.9 | |
| LN EntropyModel=Qwen2.5-32B, Number of additional samples=22026.06 | 87.9 | |
| ProbabilityModel=Gemma-3-12B, Number of additional samples=22026.06 | 87.8 | |
| Predictive EntropyModel=Qwen2.5-14B, Number of additional samples=22026.06 | 87.8 | |
| LN EntropyModel=Qwen2.5-14B, Number of additional samples=22026.06 | 87.8 | |
| PerplexityModel=Qwen3-30B-A3B, Number of additional samples=22026.06 | 87.6 | |
| P(True)Model=Qwen3-30B-A3B, Number of additional samples=22026.06 | 87.6 | |
| ProbabilityModel=Qwen3-30B-A3B, Number of additional samples=22026.06 | 87.5 | |
| Predictive EntropyModel=Gemma-3-12B, Number of additional samples=22026.06 | 87.5 | |
| Predictive EntropyModel=Gemma-3-27B, Number of additional samples=22026.06 | 87.5 | |
| LN EntropyModel=Qwen3-30B-A3B, Number of additional samples=22026.06 | 87.3 | |
| sentenceSARModel=Qwen3-30B-A3B, Number of additional samples=22026.06 | 87.2 | |
| SARModel=Qwen2.5-7B, Number of additional samples=22026.06 | 87.2 | |
| EigVModel=Gemma-3-27B, Number of additional samples=22026.06 | 87.2 | |
| DegModel=Gemma-3-27B, Number of additional samples=22026.06 | 87.2 | |
| Predictive EntropyModel=Qwen3-30B-A3B, Number of additional samples=22026.06 | 86.9 | |
| SARModel=Qwen2.5-14B, Number of additional samples=22026.06 | 86.9 | |
| SARModel=Qwen2.5-32B, Number of additional samples=22026.06 | 86.9 | |
| P(True)Model=Qwen2.5-14B, Number of additional samples=22026.06 | 86.8 | |
| tokenSARModel=Qwen3-14B, Number of additional samples=22026.06 | 86.8 | |
| SARModel=Qwen3-30B-A3B, Number of additional samples=22026.06 | 86.7 | |
| tokenSARModel=Qwen2.5-7B, Number of additional samples=22026.06 | 85.7 | |
| tokenSARModel=Qwen2.5-14B, Number of additional samples=22026.06 | 85.3 | |
| tokenSARModel=Qwen3-30B-A3B, Number of additional samples=22026.06 | 85.1 | |
| tokenSARModel=Qwen2.5-32B, Number of additional samples=22026.06 | 84.6 | |
| Semantic EntropyModel=Gemma-3-12B, Number of additional samples=22026.06 | 83.9 | |
| P(True)Model=Qwen2.5-7B, Number of additional samples=22026.06 | 83.2 | |
| Semantic EntropyModel=Gemma-3-27B, Number of additional samples=22026.06 | 83.2 | |
| EccModel=Gemma-3-27B, Number of additional samples=22026.06 | 81.7 | |
| EigVModel=Qwen3-14B, Number of additional samples=22026.06 | 81.7 | |
| EccModel=Gemma-3-12B, Number of additional samples=22026.06 | 81.2 | |
| DegModel=Qwen3-14B, Number of additional samples=22026.06 | 80.8 | |
| EigVModel=Qwen2.5-7B, Number of additional samples=22026.06 | 80.6 | |
| Eigen ScoreModel=Gemma-3-12B, Number of additional samples=22026.06 | 80.2 | |
| EigVModel=Qwen2.5-32B, Number of additional samples=22026.06 | 80.2 | |
| DegModel=Qwen2.5-7B, Number of additional samples=22026.06 | 79.8 | |
| Eigen ScoreModel=Gemma-3-27B, Number of additional samples=22026.06 | 79.5 | |
| EigVModel=Qwen3-30B-A3B, Number of additional samples=22026.06 | 79.5 | |
| DegModel=Qwen2.5-32B, Number of additional samples=22026.06 | 79.3 | |
| P(True)Model=Gemma-3-27B, Number of additional samples=22026.06 | 78.9 | |
| EigVModel=Qwen2.5-14B, Number of additional samples=22026.06 | 78.9 | |
| DegModel=Qwen3-30B-A3B, Number of additional samples=22026.06 | 78.5 | |
| EccModel=Qwen3-14B, Number of additional samples=22026.06 | 78.4 | |
| DegModel=Qwen2.5-14B, Number of additional samples=22026.06 | 78.2 | |
| Semantic EntropyModel=Qwen3-14B, Number of additional samples=22026.06 | 77.7 | |
| EccModel=Qwen2.5-7B, Number of additional samples=22026.06 | 77.3 | |
| Semantic EntropyModel=Qwen2.5-7B, Number of additional samples=22026.06 | 76.8 | |
| Eigen ScoreModel=Qwen3-14B, Number of additional samples=22026.06 | 76.7 | |
| Semantic EntropyModel=Qwen3-30B-A3B, Number of additional samples=22026.06 | 76.5 | |
| EccModel=Qwen2.5-32B, Number of additional samples=22026.06 | 76.5 | |
| Eigen ScoreModel=Qwen2.5-7B, Number of additional samples=22026.06 | 76.2 | |
| EccModel=Qwen2.5-14B, Number of additional samples=22026.06 | 75.9 | |
| Semantic EntropyModel=Qwen2.5-32B, Number of additional samples=22026.06 | 75.7 | |
| EccModel=Qwen3-30B-A3B, Number of additional samples=22026.06 | 75.7 | |
| Eigen ScoreModel=Qwen2.5-32B, Number of additional samples=22026.06 | 75.3 | |
| Semantic EntropyModel=Qwen2.5-14B, Number of additional samples=22026.06 | 74.8 | |
| Eigen ScoreModel=Qwen3-30B-A3B, Number of additional samples=22026.06 | 74.5 | |
| Eigen ScoreModel=Qwen2.5-14B, Number of additional samples=22026.06 | 74.2 | |
| P(True)Model=Gemma-3-12B, Number of additional samples=22026.06 | 73.7 |