Inference Throughput on BioASQ (Hallucination Detection)
5,351Inference Throughput (Samples/sec)Max Pooling
Evaluation Results
| Method | Links | |
|---|---|---|
| Max PoolingModel=LLaMA-3.1-8B2026.05 | 5,351 | |
| Mean PoolingModel=LLaMA-3.1-8B2026.05 | 4,649 | |
| HaMI (Original)Model=LLaMA-3.1-8B2026.05 | 4,096 | |
| Mean PoolingModel=Mistral-Nemo-Instruct (12B)2026.05 | 3,657 | |
| Max PoolingModel=Mistral-Nemo-Instruct (12B)2026.05 | 3,512 | |
| Mean PoolingModel=LLaMA-3.3-Instruct-70B2026.05 | 2,402 | |
| HaMI (Original)Model=Mistral-Nemo-Instruct (12B)2026.05 | 2,378 | |
| Max PoolingModel=LLaMA-3.3-Instruct-70B2026.05 | 2,363 | |
| HaMI (Original)Model=LLaMA-3.3-Instruct-70B2026.05 | 2,208 | |
| HaMI (SP)Model=LLaMA-3.3-Instruct-70B2026.05 | 0.19 | |
| HaMI (SP)Model=LLaMA-3.1-8B2026.05 | 0.17 | |
| HaMI (SP)Model=Mistral-Nemo-Instruct (12B)2026.05 | 0.13 |