Selective Generation on MMLU
0.947ROC-AUCTAD
Evaluation Results
| Method | Links | |
|---|---|---|
| TADBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.947 | |
| SheepsBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.945 | |
| MINDBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.939 | |
| SATRMDBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.939 | |
| LookBackLensBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.928 | |
| FactoscopeBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.909 | |
| TADBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.888 | |
| MINDBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.884 | |
| SheepsBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.881 | |
| SATRMDBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.879 | |
| LookBackLensBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.874 | |
| TADBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.855 | |
| MINDBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.847 | |
| MSPBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.846 | |
| SATRMDBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.846 | |
| Mean Token EntropyBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.841 | |
| PerplexityBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.84 | |
| Simple FocusBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.838 | |
| FocusBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.83 | |
| LookBackLensBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.828 | |
| FactoscopeBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.82 | |
| CCPBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.816 | |
| MSPBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.813 | |
| SheepsBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.806 | |
| SARBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.795 | |
| Simple FocusBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.772 | |
| MSPBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.771 | |
| Semantic EntropyBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.768 | |
| Simple FocusBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.758 | |
| LUQBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.745 | |
| Lexical Similarity Rouge-LBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.744 | |
| PerplexityBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.741 | |
| Ecc. NLI Score entail.Backbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.739 | |
| SAPLMABackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.738 | |
| EVL NLI Score entail.Backbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.735 | |
| SAPLMABase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.733 | |
| EigenScoreBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.725 | |
| DegMat NLI Score entail.Backbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.724 | |
| FactoscopeBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.718 | |
| SAPLMABackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.709 | |
| CCPBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.709 | |
| PerplexityBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.697 | |
| Semantic DensityBackbone=Gemma-2 9b, Reference Metric=Acc.2024.08 | 0.697 | |
| CCPBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.673 | |
| LUQBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.673 | |
| SARBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.663 | |
| SARBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.661 | |
| Lexical Similarity Rouge-LBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.654 | |
| FocusBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.642 | |
| EigenScoreBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.64 | |
| Ecc. NLI Score entail.Backbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.637 | |
| LUQBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.637 | |
| DegMat NLI Score entail.Base Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.633 | |
| DegMat NLI Score entail.Backbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.631 | |
| Lexical Similarity Rouge-LBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.63 | |
| EVL NLI Score entail.Base Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.63 | |
| EVL NLI Score entail.Backbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.627 | |
| Ecc. NLI Score entail.Base Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.625 | |
| Semantic DensityBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.622 | |
| Semantic DensityBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.621 | |
| Mean Token EntropyBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.619 | |
| Semantic EntropyBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.604 | |
| EigenScoreBase Model=Llama-3.1 8b, Ground Truth Metric=Acc.2024.08 | 0.603 | |
| Semantic EntropyBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.6 | |
| Mean Token EntropyBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.59 | |
| FocusBackbone=Qwen-2.5 7b, UQ Level=Sequence-level, Reference Metric=Acc.2024.08 | 0.439 |