notesum.ai
Published at October 21Who's Who: Large Language Models Meet Knowledge Conflicts in Practice
cs.SE
cs.AI
cs.LO
cs.PL
Released Date: October 21, 2024
Authors: Quang Hieu Pham1, Hoang Ngo1, Anh Tuan Luu2, Dat Quoc Nguyen1
Aff.: 1VinAI Research, Vietnam; 2Nanyang Technological University, Singapore

| Model | #Param | SimQA | W/oS | W/S |
|---|---|---|---|---|
| Gemma 1.1 | 7B | 91.8 | 12.7 | 34.4 |
| Mistral | 7B | 96.5 | 44.5 | 54.7 |
| Qwen1.5 Chat | 7B | 91.9 | 22.0 | 50.2 |
| Llama 3 | 8B | 98.2 | 72.4 | 71.5 |
| Qwen1.5 Chat | 14B | 95.5 | 9.90 | 51.9 |
| Qwen1.5 Chat | 32B | 95.5 | 21.0 | 60.5 |
| Command R | 35B | 92.5 | 69.9 | - |
| Mixtral 8x7B | 40B | 97.1 | 53.7 | 64.2 |
| Llama 3 | 70B | 97.8 | 83.8 | 86.1 |
| Qwen1.5 Chat | 72B | 95.9 | 28.2 | 66.8 |
| gpt-3.5-turbo | - | 97.9 | 36.0 | 58.9 |