Retrieval-Augmented Generation on DelucionQA
82.9FaithfulnessHyTE-Hbert
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| HyTE-HbertRetrieval Context Window Size=5, Generator Model=Llama-3.1-8B-Instruct, Judge Model=Llama-3.1-8B-Instruct2026.02 | 82.9 | 96.5 | 87.1 | |
| ModernBert*Retrieval Context Window Size=5, Generator Model=Llama-3.1-8B-Instruct, Judge Model=Llama-3.1-8B-Instruct2026.02 | 65.5 | 66.57 | 51.83 | |
| GTERetrieval Context Window Size=5, Generator Model=Llama-3.1-8B-Instruct, Judge Model=Llama-3.1-8B-Instruct2026.02 | 64.8 | 72.5 | 54.9 | |
| GemmaRetrieval Context Window Size=5, Generator Model=Llama-3.1-8B-Instruct, Judge Model=Llama-3.1-8B-Instruct2026.02 | 61.2 | 64.3 | 70.5 | |
| KaLM-mini-v1Retrieval Context Window Size=5, Generator Model=Llama-3.1-8B-Instruct, Judge Model=Llama-3.1-8B-Instruct2026.02 | 55.3 | 58.1 | 57.3 |