Open Generation on NewsRec-Chat Open Generation (test)
12.4L1 MatchGPT-4 + Hybrid
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GPT-4 + HybridBackbone=GPT-4, Retriever=Hybrid2026.05 | 12.4 | 0.5 | 18.8 | 0 | |
| NewsRec-ChatBackbone=Qwen2.5-7B-Instruct, SID Training=true2026.05 | 12.4 | 1 | 20 | 0 | |
| Qwen-7B + DenseBackbone=Qwen-7B, Retriever=Dense2026.05 | 11.8 | 0.8 | 17.4 | 0 | |
| GPT-4 + DenseBackbone=GPT-4, Retriever=Dense2026.05 | 11.8 | 0.3 | 18.3 | 0 | |
| Hist-Pop2026.05 | 11.6 | 0.7 | 16.8 | — | |
| Qwen-7B + HybridBackbone=Qwen-7B, Retriever=Hybrid2026.05 | 11.4 | 0.5 | 18.1 | 0 | |
| GPT-4 + BM25Backbone=GPT-4, Retriever=BM252026.05 | 11 | 0.5 | 15.6 | 0 | |
| BM25Retriever=BM252026.05 | 10.5 | 0.5 | 15.7 | 0 | |
| Qwen-7B + BM25Backbone=Qwen-7B, Retriever=BM252026.05 | 10.1 | 0.5 | 15.2 | 0 | |
| Popular2026.05 | 7.7 | 0.5 | 12.6 | — | |
| Random2026.05 | 5.1 | 0.1 | 10.3 | — | |
| Qwen-7B DirectBackbone=Qwen-7B, SID Training=false2026.05 | 2.4 | 0 | 6.9 | 70 | |
| GPT-4 DirectBackbone=GPT-4, SID Training=false2026.05 | 0.9 | 0 | 1.4 | 94.6 |