Recommendation on Goodbooks 10k
81.8LtCFairLRM
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| FairLRMLLM=Qwen-max, Interaction Threshold=50%2026.01 | 81.8 | 18.2 | 3.5 | 71.5 | |
| DiversityLLM=Qwen-max, Interaction Threshold=50%2026.01 | 81.3 | 19.8 | 0.9 | 69 | |
| Pop.DebiasingLLM=Qwen-max, Interaction Threshold=50%2026.01 | 81 | 18.8 | 2.3 | 69.7 | |
| Pop.DebiasingLLM=Qwen-max, Interaction Threshold=80%2026.01 | 80.8 | 18.9 | 1.7 | 68.2 | |
| FairLRMLLM=Qwen-max, Interaction Threshold=80%2026.01 | 79.8 | 17.5 | 2.9 | 74.5 | |
| FairLRMLLM=Llama-7B, User interaction threshold=822026.01 | 79.5 | 16.6 | 48 | 82.2 | |
| DiversityLLM=Qwen-max, Interaction Threshold=80%2026.01 | 79.5 | 19.1 | 1.6 | 70.5 | |
| FairLRMLLM=Llama-7B, User interaction threshold=552026.01 | 73.8 | 16.8 | 33 | 82 | |
| VanillaLLM=Qwen-max2026.01 | 69.6 | 27.4 | 2.6 | 6.4 | |
| DiversityLLM=Llama-7B, User interaction threshold=552026.01 | 62.8 | 20.5 | 14.4 | 81.2 | |
| DiversityLLM=Llama-7B, User interaction threshold=822026.01 | 51.3 | 21.4 | 22.6 | 67.9 | |
| Pop.DebiasingLLM=Llama-7B, User interaction threshold=552026.01 | 50.3 | 18.3 | 42.5 | 81.6 | |
| Pop.DebiasingLLM=Llama-7B, User interaction threshold=822026.01 | 44.9 | 23.4 | 42.5 | 80.6 | |
| VanillaLLM=Llama-7B2026.01 | 31.9 | 19.8 | 34.1 | 80.9 |