Hallucination Detection on NQ (test)
95.2AUC ROCCDES
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| CDESLLM Backbone=Llama-2-13B2025.05 | 95.2 | — | — | — | — | — | — | |
| SDESLLM Backbone=Llama-2-7B2025.05 | 94.1 | — | — | — | — | — | — | |
| CDESLLM Backbone=Llama-2-7B2025.05 | 93.7 | — | — | — | — | — | — | |
| MINDLLM Backbone=Llama-2-13B2025.05 | 93.6 | — | — | — | — | — | — | |
| CDESLLM Backbone=Vicuna-13B2025.05 | 93.6 | — | — | — | — | — | — | |
| SDESLLM Backbone=Alpaca-13B2025.05 | 93.5 | — | — | — | — | — | — | |
| SAPLMALLM Backbone=Vicuna-13B2025.05 | 93.2 | — | — | — | — | — | — | |
| SDESLLM Backbone=Llama-2-13B2025.05 | 93.2 | — | — | — | — | — | — | |
| Probe@ExactLLM Backbone=Alpaca-13B2025.05 | 92.8 | — | — | — | — | — | — | |
| MINDLLM Backbone=Alpaca-13B2025.05 | 92.7 | — | — | — | — | — | — | |
| ODESLLM Backbone=Vicuna-13B2025.05 | 92.4 | — | — | — | — | — | — | |
| Probe@ExactLLM Backbone=Vicuna-13B2025.05 | 92.3 | — | — | — | — | — | — | |
| ODESLLM Backbone=Llama-2-13B2025.05 | 92.1 | — | — | — | — | — | — | |
| CDESLLM Backbone=Alpaca-13B2025.05 | 92.1 | — | — | — | — | — | — | |
| Probe@ExactLLM Backbone=Llama-2-7B2025.05 | 92 | — | — | — | — | — | — | |
| Probe@ExactLLM Backbone=Llama-2-13B2025.05 | 91.9 | — | — | — | — | — | — | |
| CDESLLM Backbone=Gemma-2-9B2025.05 | 91.8 | — | — | — | — | — | — | |
| ODESLLM Backbone=Llama-2-7B2025.05 | 91.3 | — | — | — | — | — | — | |
| SDESLLM Backbone=Vicuna-13B2025.05 | 91.1 | — | — | — | — | — | — | |
| SDESLLM Backbone=Gemma-2-9B2025.05 | 90.9 | — | — | — | — | — | — | |
| MINDLLM Backbone=Vicuna-13B2025.05 | 90.6 | — | — | — | — | — | — | |
| SAPLMALLM Backbone=Llama-2-13B2025.05 | 90.5 | — | — | — | — | — | — | |
| MINDLLM Backbone=Llama-2-7B2025.05 | 90.5 | — | — | — | — | — | — | |
| ODESLLM Backbone=Alpaca-13B2025.05 | 90.5 | — | — | — | — | — | — | |
| CDESLLM Backbone=Mistral-7B-v0.32025.05 | 90.5 | — | — | — | — | — | — | |
| SAPLMALLM Backbone=Mistral-7B-v0.32025.05 | 90.3 | — | — | — | — | — | — | |
| Probe@ExactLLM Backbone=Gemma-2-9B2025.05 | 90.3 | — | — | — | — | — | — | |
| ODESLLM Backbone=Gemma-2-9B2025.05 | 90 | — | — | — | — | — | — | |
| ODESLLM Backbone=Mistral-7B-v0.32025.05 | 89.7 | — | — | — | — | — | — | |
| SDESLLM Backbone=Mistral-7B-v0.32025.05 | 89.7 | — | — | — | — | — | — | |
| MINDLLM Backbone=Gemma-2-9B2025.05 | 89.5 | — | — | — | — | — | — | |
| SAPLMALLM Backbone=Llama-2-7B2025.05 | 89.1 | — | — | — | — | — | — | |
| SAPLMALLM Backbone=Gemma-2-9B2025.05 | 88.9 | — | — | — | — | — | — | |
| Probe@ExactLLM Backbone=Mistral-7B-v0.32025.05 | 88.6 | — | — | — | — | — | — | |
| SAPLMALLM Backbone=Alpaca-13B2025.05 | 87.6 | — | — | — | — | — | — | |
| MINDLLM Backbone=Mistral-7B-v0.32025.05 | 87.2 | — | — | — | — | — | — | |
| TDGNetgraph_type=dynamic, decoding_gen_length=32, decoding_steps=16, granularity=token-level2026.02 | 87 | — | — | — | — | — | — | |
| CHARMgraph_type=static, decoding_gen_length=32, decoding_steps=16, granularity=token-level2026.02 | 84 | — | — | — | — | — | — | |
| ACSELLM Backbone=Mistral-7B-Instruct2026.05 | 84 | — | — | 36 | 29 | 81 | 87 | |
| CAPLLM Backbone=Mistral-7B-Instruct2026.05 | 80 | — | — | 50 | 42 | 78 | 77 | |
| EUBHDLLM Backbone=Llama-2-13B2025.05 | 79.4 | — | — | — | — | — | — | |
| DDCRP-CPLLM Backbone=Mistral-7B-Instruct2026.05 | 78 | — | — | 54 | 46 | 75 | 75 | |
| INSIDELLM Backbone=Llama-2-13B2025.05 | 77.1 | — | — | — | — | — | — | |
| EUBHDLLM Backbone=Alpaca-13B2025.05 | 76.8 | — | — | — | — | — | — | |
| INSIDELLM Backbone=Llama-2-7B2025.05 | 76.8 | — | — | — | — | — | — | |
| INSIDELLM Backbone=Mistral-7B-v0.32025.05 | 76.4 | — | — | — | — | — | — | |
| INSIDELLM Backbone=Gemma-2-9B2025.05 | 75.9 | — | — | — | — | — | — | |
| INSIDELLM Backbone=Vicuna-13B2025.05 | 75.8 | — | — | — | — | — | — | |
| SULLM Backbone=Mistral-7B-Instruct2026.05 | 75 | — | — | 58 | 50 | 72 | 73 | |
| INSIDELLM Backbone=Alpaca-13B2025.05 | 74.3 | — | — | — | — | — | — | |
| EigVLLM Backbone=Mistral-7B-Instruct2026.05 | 74 | — | — | 66 | 59 | 70 | 69 | |
| EUBHDLLM Backbone=Llama-2-7B2025.05 | 73.9 | — | — | — | — | — | — | |
| EUBHDLLM Backbone=Vicuna-13B2025.05 | 73.2 | — | — | — | — | — | — | |
| EUBHDLLM Backbone=Mistral-7B-v0.32025.05 | 71.7 | — | — | — | — | — | — | |
| EUBHDLLM Backbone=Gemma-2-9B2025.05 | 70.3 | — | — | — | — | — | — | |
| Predictive Entropydecoding_gen_length=32, decoding_steps=16, granularity=token-level2026.02 | 69 | — | — | — | — | — | — | |
| UnigramLLM Backbone=Vicuna-13B2025.05 | 67.8 | — | — | — | — | — | — | |
| P(True)LLM Backbone=Mistral-7B-Instruct2026.05 | 67 | — | — | 74 | 66 | 64 | 64 | |
| NLILLM Backbone=Llama-2-13B2025.05 | 66.9 | — | — | — | — | — | — | |
| NLILLM Backbone=Vicuna-13B2025.05 | 65.4 | — | — | — | — | — | — | |
| UnigramLLM Backbone=Llama-2-13B2025.05 | 65.2 | — | — | — | — | — | — | |
| UnigramLLM Backbone=Gemma-2-9B2025.05 | 65.1 | — | — | — | — | — | — | |
| UnigramLLM Backbone=Mistral-7B-v0.32025.05 | 64.5 | — | — | — | — | — | — | |
| NLILLM Backbone=Llama-2-7B2025.05 | 64.2 | — | — | — | — | — | — | |
| NLILLM Backbone=Gemma-2-9B2025.05 | 64 | — | — | — | — | — | — | |
| NLILLM Backbone=Alpaca-13B2025.05 | 63.8 | — | — | — | — | — | — | |
| UnigramLLM Backbone=Llama-2-7B2025.05 | 63.1 | — | — | — | — | — | — | |
| UnigramLLM Backbone=Alpaca-13B2025.05 | 62.9 | — | — | — | — | — | — | |
| NLILLM Backbone=Mistral-7B-v0.32025.05 | 62.6 | — | — | — | — | — | — | |
| TELLM Backbone=Mistral-7B-Instruct2026.05 | 62 | — | — | 82 | 75 | 57 | 60 | |
| Token Probabilitydecoding_gen_length=32, decoding_steps=16, granularity=token-level2026.02 | 60 | — | — | — | — | — | — | |
| P(True)LLM Backbone=Llama-2-13B2025.05 | 56.8 | — | — | — | — | — | — | |
| AvgProbLLM Backbone=Vicuna-13B2025.05 | 56.4 | — | — | — | — | — | — | |
| AvgProbLLM Backbone=Llama-2-13B2025.05 | 55.9 | — | — | — | — | — | — | |
| AvgProbLLM Backbone=Gemma-2-9B2025.05 | 55.3 | — | — | — | — | — | — | |
| AvgEntLLM Backbone=Vicuna-13B2025.05 | 55.2 | — | — | — | — | — | — | |
| P(True)LLM Backbone=Llama-2-7B2025.05 | 54.7 | — | — | — | — | — | — | |
| AvgProbLLM Backbone=Mistral-7B-v0.32025.05 | 54.7 | — | — | — | — | — | — | |
| AvgEntLLM Backbone=Llama-2-13B2025.05 | 54.6 | — | — | — | — | — | — | |
| AvgEntLLM Backbone=Gemma-2-9B2025.05 | 54.6 | — | — | — | — | — | — | |
| AvgProbLLM Backbone=Llama-2-7B2025.05 | 54.3 | — | — | — | — | — | — | |
| AvgEntLLM Backbone=Alpaca-13B2025.05 | 54.2 | — | — | — | — | — | — | |
| AvgEntLLM Backbone=Llama-2-7B2025.05 | 53.9 | — | — | — | — | — | — | |
| AvgEntLLM Backbone=Mistral-7B-v0.32025.05 | 53.8 | — | — | — | — | — | — | |
| P(True)LLM Backbone=Vicuna-13B2025.05 | 53.4 | — | — | — | — | — | — | |
| AvgProbLLM Backbone=Alpaca-13B2025.05 | 53.1 | — | — | — | — | — | — | |
| P(True)LLM Backbone=Mistral-7B-v0.32025.05 | 52.1 | — | — | — | — | — | — | |
| Source Attributiondecoding_gen_length=32, decoding_steps=16, granularity=token-level2026.02 | 52 | — | — | — | — | — | — | |
| P(True)LLM Backbone=Gemma-2-9B2025.05 | 51.3 | — | — | — | — | — | — | |
| P(True)LLM Backbone=Alpaca-13B2025.05 | 51 | — | — | — | — | — | — | |
| Graph Degree (Centrality)decoding_gen_length=32, decoding_steps=16, granularity=token-level2026.02 | 43 | — | — | — | — | — | — | |
| CRGn_cal=595, Target=95%2025.12 | — | 95.8 | 0 | — | — | — | — |