Hallucination Detection on SQuAD In-domain (test)
92.2AccuracyAURORA
Evaluation Results
| Method | Links | |
|---|---|---|
| AURORALLM Backbone=Gemma3-12B-it2026.06 | 92.2 | |
| HARPLLM Backbone=Gemma3-12B-it2026.06 | 91.95 | |
| HARPLLM Backbone=Ministral3-8B-Base2026.06 | 89.25 | |
| HARPLLM Backbone=Llama-3-8B-it2026.06 | 87.45 | |
| HARPLLM Backbone=Qwen3-8B-it2026.06 | 87.05 | |
| AURORALLM Backbone=Llama-3-8B-it2026.06 | 85.2 | |
| AURORALLM Backbone=Qwen3-8B-it2026.06 | 85.2 | |
| AURORALLM Backbone=Ministral3-8B-Base2026.06 | 85.05 | |
| ICR probeLLM Backbone=Llama-3-8B-it2026.06 | 78.8 | |
| ICR probeLLM Backbone=Ministral3-8B-Base2026.06 | 76.75 | |
| ICR probeLLM Backbone=Gemma3-12B-it2026.06 | 73.25 | |
| InterrogateLLMLLM Backbone=Qwen3-8B-it2026.06 | 72.8 | |
| InterrogateLLMLLM Backbone=Llama-3-8B-it2026.06 | 69.6 | |
| ICR probeLLM Backbone=Qwen3-8B-it2026.06 | 68.15 | |
| InterrogateLLMLLM Backbone=Gemma3-12B-it2026.06 | 67.2 | |
| InterrogateLLMLLM Backbone=Ministral3-8B-Base2026.06 | 64.2 |