Topic Ranking Evaluation on Wiki (test)
0.48Kendall's τ (Fit)PROXANN
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PROXANNBackbone=Qwen-2.5-72B2025.07 | 0.48 | 0.31 | |
| PROXANNBackbone=Llama-3.1-70B2025.07 | 0.41 | 0.36 | |
| HumanProtocol=Leave-one-out inter-annotator correlation2025.07 | 0.41 | 0.34 | |
| PROXANNBackbone=Qwen-2-7B2025.07 | 0.35 | 0.33 | |
| PROXANNBackbone=GPT-4o2025.07 | 0.22 | 0.27 | |
| PROXANNBackbone=Qwen-2.5-32B2025.07 | 0.2 | 0.51 | |
| PROXANNBackbone=Llama-3.1-8B2025.07 | 0.19 | -0.35 | |
| NPMI2025.07 | -0.15 | -0.18 |