Query Recommendation on Amazon Public (Offline)
12.3Q_EMGemini-3.1-pro
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Gemini-3.1-proMethod Category=LLM Direct Inference2026.06 | 12.3 | 9.1 | 2.1 | |
| Gemini-3.0-proMethod Category=LLM Direct Inference2026.06 | 11.5 | 8 | 1.9 | |
| Qwen3.6-plusMethod Category=LLM Direct Inference2026.06 | 9.6 | 5.8 | 1.5 | |
| QueryAgent-R1Method Category=Ours2026.06 | 9.5 | 14.4 | 6.3 | |
| +RetrievalMethod Category=Ours, Base Model=Qwen3-4B2026.06 | 8.3 | 7.7 | 1.3 | |
| DeepSeek-v4-proMethod Category=LLM Direct Inference2026.06 | 8.2 | 9.1 | 1.1 | |
| DeepSeek-v4-flashMethod Category=LLM Direct Inference2026.06 | 7 | 9.1 | 1.3 | |
| +RL trainingMethod Category=Ours, Base Model=Qwen3-4B2026.06 | 5.7 | 6.2 | 0.9 | |
| Qwen3-Emb-4BMethod Category=Inventory-based Retrieval2026.06 | 4.1 | 4.2 | 0.2 | |
| Qwen3-4B BackboneMethod Category=Ours2026.06 | 4.1 | 4.2 | 0.2 | |
| GPT-5.1Method Category=LLM Direct Inference2026.06 | 3.9 | 4.2 | 0.2 | |
| SwingMethod Category=Inventory-based Retrieval2026.06 | 3.7 | 0.6 | 0.3 | |
| Qwen3-MaxMethod Category=LLM Direct Inference2026.06 | 3.5 | 4.7 | 0.1 | |
| Qwen3-Emb-0.6BMethod Category=Inventory-based Retrieval2026.06 | 3.3 | 0.8 | 0.3 | |
| GPT-4oMethod Category=LLM Direct Inference2026.06 | 3.3 | 3 | 0.2 |