Biomedical Question Answering on BioASQ
29Factoid AccLLAMA-70B+7B (PT)
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| LLAMA-70B+7B (PT)Model Training Status=Finetuned, Collaboration Strategy=Proxy Tuning2024.03 | 29 | 16.8 | 85.2 | 21.3 | 38.1 | — | — | |
| Co-GreedyCollaboration Strategy=Greedy2024.03 | 29 | 16.6 | 76.5 | 20.2 | 35.6 | — | — | |
| LLAMA-70B (QLORA)Model Training Status=Finetuned, Strategy=QLoRA2024.03 | 24.7 | 20.7 | 75.3 | 21.1 | 35.5 | — | — | |
| Co-LLM-7B + LLAMA-70BCollaboration Strategy=Co-LLM2024.03 | 24.7 | 18 | 82.7 | 20.4 | 36.5 | — | — | |
| LLAMA-7B (Finetuned)Model Training Status=Finetuned2024.03 | 23.7 | 13.8 | 76.5 | 18.1 | 33 | — | — | |
| Weak SupervisionCollaboration Strategy=Weak Supervision2024.03 | 22.6 | 14.6 | 80.2 | 17.5 | 33.7 | — | — | |
| Co-LLM-7B (Base Only)Collaboration Strategy=Co-LLM (Base Only)2024.03 | 20.4 | 11.2 | 79 | 18.1 | 32.2 | — | — | |
| LLAMA-70BModel Training Status=Untuned, Prompting Strategy=1-shot2024.03 | 11.8 | 14.9 | 77.8 | 18.6 | 30.8 | — | — | |
| LLAMA-70B+7B (CD)Model Training Status=Untuned, Collaboration Strategy=Contrastive Decoding, Prompting Strategy=1-shot2024.03 | 11.8 | 9 | 71.6 | 17.5 | 27.5 | — | — | |
| Co-RandomCollaboration Strategy=Random2024.03 | 6.5 | 1.9 | 30.9 | 17.5 | 14.3 | — | — | |
| LLAMA-7BModel Training Status=Untuned, Prompting Strategy=1-shot2024.03 | 4.3 | 4.9 | 71.6 | 17.2 | 24.5 | — | — |