Hallucination Detection on Hallucination-detection dataset
86.3AUROCneuron-based linear probe
Evaluation Results
| Method | Links | |
|---|---|---|
| neuron-based linear probeBackbone=Llama-3.12026.06 | 86.3 | |
| neuron-based linear probeBackbone=BioMistral2026.06 | 80.8 | |
| neuron-based linear probeBackbone=Meditron2026.06 | 79.6 | |
| saplmaBackbone=Llama-3.12026.06 | 79.16 | |
| saplmaBackbone=Meditron2026.06 | 77.29 | |
| neuron-based linear probeBackbone=PMC-LLaMA2026.06 | 76.7 | |
| token_neg_probBackbone=Llama-3.12026.06 | 76.44 | |
| haloscopeBackbone=Meditron2026.06 | 75.63 | |
| ccsBackbone=Meditron2026.06 | 75.18 | |
| saplmaBackbone=BioMistral2026.06 | 74.69 | |
| haloscopeBackbone=PMC-LLaMA2026.06 | 74.53 | |
| ccsBackbone=PMC-LLaMA2026.06 | 73.76 | |
| token_neg_probBackbone=BioMistral2026.06 | 73.58 | |
| saplmaBackbone=PMC-LLaMA2026.06 | 73.09 | |
| haloscopeBackbone=BioMistral2026.06 | 72.43 | |
| ccsBackbone=BioMistral2026.06 | 71.54 | |
| token_entropyBackbone=Llama-3.12026.06 | 70.61 | |
| haloscopeBackbone=Llama-3.12026.06 | 67.71 | |
| ccsBackbone=Llama-3.12026.06 | 63.49 | |
| token_entropyBackbone=BioMistral2026.06 | 60.25 | |
| token_neg_probBackbone=Meditron2026.06 | 49.67 | |
| token_entropyBackbone=Meditron2026.06 | 30.68 | |
| token_neg_probBackbone=PMC-LLaMA2026.06 | 26.63 | |
| token_entropyBackbone=PMC-LLaMA2026.06 | 26.3 |