Retrieval on Email Dataset
77.9Hit@1Qwen3-8B + DAPO
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen3-8B + DAPOCategory=DSL-R1, Backbone=Qwen3-8B, RL Strategy=DAPO2026.01 | 77.9 | 94.2 | 86.1 | 78 | |
| Qwen3-8B + GRPOCategory=DSL-R1, Backbone=Qwen3-8B, RL Strategy=GRPO2026.01 | 75.8 | 92.9 | 84.6 | 82 | |
| Qwen3-4B + DAPOCategory=DSL-R1, Backbone=Qwen3-4B, RL Strategy=DAPO2026.01 | 71.7 | 90.6 | 80.9 | 52 | |
| Qwen3-4B + GRPOCategory=DSL-R1, Backbone=Qwen3-4B, RL Strategy=GRPO2026.01 | 69.5 | 88.9 | 78.6 | 55 | |
| GPT-4oCategory=LLM Based2026.01 | 68.9 | 88.7 | 78.8 | 30 | |
| Qwen3-8BCategory=LLM Based2026.01 | 62.1 | 83.6 | 72.5 | 140 | |
| Qwen3-4BCategory=LLM Based2026.01 | 58.3 | 80.1 | 69.2 | 95 | |
| ColQwenCategory=Traditional Retrieval2026.01 | 55.3 | 78.9 | 65.4 | 35 | |
| BM25Category=Traditional Retrieval2026.01 | 42.1 | 64.7 | 53.2 | 18 |