Conversational Music Recommendation on TalkPlay (test)
0.049MRRTALKPLAY
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| TALKPLAYRetrieval Method=Generative, Query Understanding=true, Response Generation=true, Protocol=Finetuned2025.02 | 0.049 | 2.6 | 28.8 | |
| NV-Embed-V2Retrieval Method=Similarity-based, Query Understanding=true, Response Generation=false, Protocol=Finetuned2025.02 | 0.032 | 0.5 | 28.1 | |
| BM25Retrieval Method=Sparse, Query Understanding=true, Response Generation=false, Protocol=Zero-shot2025.02 | 0.03 | 0.2 | 26.5 | |
| NV-Embed-V2Retrieval Method=Similarity-based, Query Understanding=true, Response Generation=false, Protocol=Zero-shot2025.02 | 0.029 | 0.5 | 25.5 | |
| CLAP-MusicRetrieval Method=Similarity-based, Query Understanding=true, Response Generation=false, Protocol=Finetuned2025.02 | 0.017 | 0.5 | 18.9 | |
| CLAP-MusicRetrieval Method=Similarity-based, Query Understanding=true, Response Generation=false, Protocol=Zero-shot2025.02 | 0.011 | 0.3 | 15.3 | |
| TIGERRetrieval Method=Generative, Query Understanding=false, Response Generation=false, Protocol=Finetuned2025.02 | 0.008 | 0.1 | 9.2 | |
| SASRecRetrieval Method=Generative, Query Understanding=false, Response Generation=false, Protocol=Finetuned2025.02 | 0.002 | 0 | 0.8 |