Open-ended Generation on TriviaQA
5.18ECEActCab
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ActCabBackbone=Llama2-7B2025.09 | 5.18 | 23.75 | 56.21 | — | — | |
| GrACEBackbone=Llama3.1-8B2025.09 | 5.94 | 12.95 | 81.03 | 78.4 | — | |
| UQ4CTBackbone=Llama-3.1-8B-Instruct, Fine-tuning=MixLoRA, Number of samples (N)=1, Correctness proxy=F1-based2024.10 | 6.63 | — | — | 72.4 | — | |
| GrACEBackbone=Phi3-3.8B2025.09 | 7.63 | 16.64 | 81.85 | 64.7 | — | |
| ActCabBackbone=Llama3.1-8B2025.09 | 7.84 | 16.01 | 72.74 | — | — | |
| GrACEBackbone=Llama2-7B2025.09 | 8.41 | 16.78 | 84.22 | 59.1 | — | |
| ActCabBackbone=Phi3-3.8B2025.09 | 8.47 | 22.46 | 64.27 | — | — | |
| Semantic EntropyBackbone=Llama-3.1-8B-Instruct, Fine-tuning=MixLoRA, Number of samples (N)=5, Correctness proxy=F1-based2024.10 | 8.56 | — | — | 71.82 | — | |
| ApricotBackbone=Phi3-3.8B2025.09 | 9.69 | 17.85 | 78.36 | — | — | |
| ApricotBackbone=Llama3.1-8B2025.09 | 9.8 | 15.56 | 74.44 | — | — | |
| P(True)Backbone=Llama2-7B2025.09 | 9.82 | 23.12 | 66.02 | — | — | |
| Platt ScalingBackbone=Phi3-3.8B2025.09 | 10.75 | 22.43 | 81.45 | — | — | |
| Verbalized ConsistencyBackbone=Llama-3.1-8B-Instruct, Fine-tuning=MixLoRA, Number of samples (N)=5, Correctness proxy=F1-based2024.10 | 11.01 | — | — | 72.08 | — | |
| VerbalBackbone=Llama3.1-8B2025.09 | 11.45 | 17.36 | 69.96 | — | — | |
| ApricotBackbone=Llama2-7B2025.09 | 11.88 | 18.55 | 81.61 | — | — | |
| Seq. LikelihoodBackbone=Llama2-7B2025.09 | 12.56 | 21.88 | 73.27 | 59.9 | — | |
| P(True)Backbone=Phi3-3.8B2025.09 | 12.65 | 17.85 | 81.69 | — | — | |
| Semantic DensityBackbone=Llama-3.1-8B-Instruct, Fine-tuning=MixLoRA, Number of samples (N)=5, Correctness proxy=F1-based2024.10 | 13.1 | — | — | 71.82 | — | |
| Platt ScalingBackbone=Llama3.1-8B2025.09 | 13.26 | 16.96 | 79.44 | — | — | |
| UA-CLMBackbone=Llama-2-13B2024.12 | 13.65 | — | — | 77.1 | 72.77 | |
| Platt ScalingBackbone=Llama2-7B2025.09 | 13.95 | 23.47 | 73.27 | — | — | |
| Seq. LikelihoodBackbone=Phi3-3.8B2025.09 | 14.57 | 20.62 | 81.45 | 64.4 | — | |
| ULTBackbone=Llama-2-13B2024.12 | 15.17 | — | — | 73.09 | 68.75 | |
| Seq. LikelihoodBackbone=Llama3.1-8B2025.09 | 15.71 | 17.32 | 79.44 | 78 | — | |
| P(True)Backbone=Llama3.1-8B2025.09 | 17.33 | 22.24 | 53.73 | — | — | |
| SeqProbBackbone=Llama-3.1-8B-Instruct, Fine-tuning=MixLoRA, Number of samples (N)=1, Correctness proxy=F1-based2024.10 | 19.23 | — | — | 71.61 | — | |
| CTBackbone=Llama-2-13B2024.12 | 19.37 | — | — | 74.29 | 70.18 | |
| VerbalizedBackbone=Llama-3.1-8B-Instruct, Fine-tuning=MixLoRA, Number of samples (N)=1, Correctness proxy=F1-based2024.10 | 20.54 | — | — | 72.05 | — | |
| UA-CLMBackbone=Llama-2-7B2024.12 | 20.9 | — | — | 71.08 | 66.79 | |
| Pre-trainedBackbone=Llama-2-13B2024.12 | 21.33 | — | — | 76.1 | 71.6 | |
| CLMBackbone=Llama-2-13B2024.12 | 22.41 | — | — | 69.67 | 65.88 | |
| CTBackbone=Llama-2-7B2024.12 | 22.76 | — | — | 69.87 | 66 | |
| Pre-trainedBackbone=Llama-2-7B2024.12 | 23.04 | — | — | 70.48 | 66.54 | |
| CLMBackbone=Llama-2-7B2024.12 | 24.07 | — | — | 65.29 | 60.37 | |
| VerbalBackbone=Phi3-3.8B2025.09 | 27.34 | 28.96 | 63.01 | — | — | |
| ULTBackbone=Llama-2-7B2024.12 | 31.11 | — | — | 65.86 | 61.21 | |
| VerbalBackbone=Llama2-7B2025.09 | 34.31 | 37.02 | 50.89 | — | — | |
| ADModel=LLaMA3-8B, Decoding=AD2026.05 | — | — | — | 70.48 | — | |
| ADModel=LLaMA2-7B, Decoding=AD2026.05 | — | — | — | 63.57 | — | |
| DeLaskModel=LLaMA3-8B, Decoding=DeLask2026.05 | — | — | — | 72.65 | — | |
| DeLaskModel=LLaMA2-7B, Decoding=DeLask2026.05 | — | — | — | 64.32 | — | |
| DOLAModel=LLaMA3-8B, Decoding=DOLA2026.05 | — | — | — | 70.77 | — | |
| DOLAModel=LLaMA2-7B, Decoding=DOLA2026.05 | — | — | — | 62.5 | — | |
| ITIModel=LLaMA3-8B, Decoding=ITI2026.05 | — | — | — | 67.37 | — | |
| ITIModel=LLaMA2-7B, Decoding=ITI2026.05 | — | — | — | 63.89 | — | |
| RegularModel=LLaMA3-8B, Decoding=Regular2026.05 | — | — | — | 71.61 | — | |
| RegularModel=LLaMA2-7B, Decoding=Regular2026.05 | — | — | — | 63.34 | — | |
| UADModel=LLaMA3-8B, Decoding=UAD2026.05 | — | — | — | 64.09 | — | |
| UADModel=LLaMA2-7B, Decoding=UAD2026.05 | — | — | — | 54.09 | — |