Question Answering on NQ (test) (AUROC)
83AUROCCCP
Evaluation results
| Method | Links | |
|---|---|---|
| CCPLLM Backbone=Qwen2.5-7B2026.03 | 83 | |
| Semantic Token ClusteringLLM Backbone=Qwen2.5-7B2026.03 | 82.7 | |
| EigVLLM Backbone=Qwen2.5-7B2026.03 | 82.6 | |
| Semantic Token ClusteringLLM Backbone=Qwen2.5-3B2026.03 | 81.7 | |
| Semantic EntropyLLM Backbone=Qwen2.5-7B2026.03 | 81.1 | |
| EigVLLM Backbone=Qwen2.5-3B2026.03 | 80.8 | |
| CCPLLM Backbone=Qwen2.5-3B2026.03 | 80.4 | |
| CCPLLM Backbone=Qwen2.5-14B2026.03 | 80.4 | |
| Semantic EntropyLLM Backbone=Qwen2.5-3B2026.03 | 80 | |
| Semantic Token ClusteringLLM Backbone=Qwen2.5-14B2026.03 | 79.8 | |
| DegLLM Backbone=Qwen2.5-7B2026.03 | 79.6 | |
| ConULLM Backbone=Qwen2.5-7B2026.03 | 78.8 | |
| SARLLM Backbone=Qwen2.5-7B2026.03 | 78.6 | |
| DegLLM Backbone=Qwen2.5-3B2026.03 | 78.5 | |
| SARLLM Backbone=Qwen2.5-3B2026.03 | 78.4 | |
| EigVLLM Backbone=Qwen2.5-14B2026.03 | 78.3 | |
| Semantic EntropyLLM Backbone=Qwen2.5-14B2026.03 | 77.9 | |
| Semantic Token ClusteringLLM Backbone=Llama-3-8B2026.03 | 77.7 | |
| Semantic Token ClusteringLLM Backbone=Mistral-7B2026.03 | 77.6 | |
| sentenceSARLLM Backbone=Qwen2.5-3B2026.03 | 77.4 | |
| sentenceSARLLM Backbone=Qwen2.5-7B2026.03 | 77.2 | |
| CCPLLM Backbone=Mistral-7B2026.03 | 76.8 | |
| Predictive EntropyLLM Backbone=Qwen2.5-3B2026.03 | 76.7 | |
| EigVLLM Backbone=Llama-3-8B2026.03 | 76.7 | |
| Semantic Token ClusteringLLM Backbone=Llama-2-7B2026.03 | 76.7 | |
| LN EntropyLLM Backbone=Qwen2.5-3B2026.03 | 76.4 | |
| Predictive EntropyLLM Backbone=Qwen2.5-7B2026.03 | 76.3 | |
| CCPLLM Backbone=Llama-3-8B2026.03 | 76.3 | |
| EigVLLM Backbone=Mistral-7B2026.03 | 76.2 | |
| CCPLLM Backbone=Llama-2-7B2026.03 | 76.1 | |
| PerplexityLLM Backbone=Qwen2.5-3B2026.03 | 76 | |
| ConULLM Backbone=Qwen2.5-3B2026.03 | 75.9 | |
| EccLLM Backbone=Qwen2.5-7B2026.03 | 75.8 | |
| DegLLM Backbone=Qwen2.5-14B2026.03 | 75.7 | |
| Semantic EntropyLLM Backbone=Llama-3-8B2026.03 | 75.6 | |
| LN EntropyLLM Backbone=Qwen2.5-7B2026.03 | 75.5 | |
| Semantic EntropyLLM Backbone=Mistral-7B2026.03 | 75.4 | |
| SARLLM Backbone=Qwen2.5-14B2026.03 | 75.4 | |
| EccLLM Backbone=Qwen2.5-3B2026.03 | 75.3 | |
| EigVLLM Backbone=Llama-2-7B2026.03 | 75.3 | |
| DegLLM Backbone=Llama-3-8B2026.03 | 75.2 | |
| Eigen ScoreLLM Backbone=Qwen2.5-3B2026.03 | 75 | |
| sentenceSARLLM Backbone=Qwen2.5-14B2026.03 | 74.9 | |
| ConULLM Backbone=Qwen2.5-14B2026.03 | 74.8 | |
| Eigen ScoreLLM Backbone=Qwen2.5-7B2026.03 | 74.7 | |
| PerplexityLLM Backbone=Qwen2.5-7B2026.03 | 74.4 | |
| Semantic EntropyLLM Backbone=Llama-2-7B2026.03 | 74.3 | |
| Predictive EntropyLLM Backbone=Qwen2.5-14B2026.03 | 74.2 | |
| DegLLM Backbone=Mistral-7B2026.03 | 74.2 | |
| sentenceSARLLM Backbone=Llama-3-8B2026.03 | 73.7 | |
| tokenSARLLM Backbone=Qwen2.5-3B2026.03 | 73.6 | |
| DegLLM Backbone=Llama-2-7B2026.03 | 73.5 | |
| SARLLM Backbone=Llama-3-8B2026.03 | 73.1 | |
| Predictive EntropyLLM Backbone=Llama-3-8B2026.03 | 73 | |
| sentenceSARLLM Backbone=Mistral-7B2026.03 | 72.8 | |
| SARLLM Backbone=Mistral-7B2026.03 | 72.8 | |
| P(True)LLM Backbone=Qwen2.5-7B2026.03 | 72.7 | |
| EccLLM Backbone=Qwen2.5-14B2026.03 | 72.7 | |
| P(True)LLM Backbone=Qwen2.5-14B2026.03 | 72.5 | |
| tokenSARLLM Backbone=Qwen2.5-7B2026.03 | 72.4 | |
| EccLLM Backbone=Llama-3-8B2026.03 | 72.3 | |
| sentenceSARLLM Backbone=Llama-2-7B2026.03 | 72.3 | |
| SARLLM Backbone=Llama-2-7B2026.03 | 72.3 | |
| Predictive EntropyLLM Backbone=Mistral-7B2026.03 | 72.2 | |
| LN EntropyLLM Backbone=Qwen2.5-14B2026.03 | 72.1 | |
| Predictive EntropyLLM Backbone=Llama-2-7B2026.03 | 71.6 | |
| Eigen ScoreLLM Backbone=Qwen2.5-14B2026.03 | 71.6 | |
| ConULLM Backbone=Mistral-7B2026.03 | 71.5 | |
| LN EntropyLLM Backbone=Mistral-7B2026.03 | 71.4 | |
| EccLLM Backbone=Mistral-7B2026.03 | 71.4 | |
| PerplexityLLM Backbone=Mistral-7B2026.03 | 71.2 | |
| LN EntropyLLM Backbone=Llama-3-8B2026.03 | 71.2 | |
| PerplexityLLM Backbone=Qwen2.5-14B2026.03 | 70.8 | |
| tokenSARLLM Backbone=Qwen2.5-14B2026.03 | 70.8 | |
| ConULLM Backbone=Llama-3-8B2026.03 | 70.8 | |
| ConULLM Backbone=Llama-2-7B2026.03 | 70.6 | |
| EccLLM Backbone=Llama-2-7B2026.03 | 70.6 | |
| PerplexityLLM Backbone=Llama-3-8B2026.03 | 70.5 | |
| LN EntropyLLM Backbone=Llama-2-7B2026.03 | 70.2 | |
| PerplexityLLM Backbone=Llama-2-7B2026.03 | 69.7 | |
| Eigen ScoreLLM Backbone=Llama-2-7B2026.03 | 69.1 | |
| Eigen ScoreLLM Backbone=Llama-3-8B2026.03 | 69.1 | |
| Eigen ScoreLLM Backbone=Mistral-7B2026.03 | 68.4 | |
| tokenSARLLM Backbone=Mistral-7B2026.03 | 68.3 | |
| tokenSARLLM Backbone=Llama-3-8B2026.03 | 68.2 | |
| P(True)LLM Backbone=Qwen2.5-3B2026.03 | 68 | |
| tokenSARLLM Backbone=Llama-2-7B2026.03 | 67.3 | |
| P(True)LLM Backbone=Llama-2-7B2026.03 | 55 | |
| P(True)LLM Backbone=Llama-3-8B2026.03 | 54.6 | |
| P(True)LLM Backbone=Mistral-7B2026.03 | 48.8 |