Product Search on Amazon Product Search ESCI
52.8NDCG@5FlexRec
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| FlexRecEvaluation Protocol=Post-Trained LLM, Base Model=Qwen2.5-3B-Instruct2026.03 | 52.8 | 55.3 | 69 | 27.7 | 67.8 | |
| Rec-R1Evaluation Protocol=Post-Trained LLM, Base Model=Qwen2.5-3B-Instruct2026.03 | 50.4 | 53.6 | 67.9 | 26.8 | 65.2 | |
| GPT-4oEvaluation Protocol=Zero-Shot LLM2026.03 | 50.2 | 53 | 67.5 | 26.4 | 64.7 | |
| Qwen2.5-3B-InstructEvaluation Protocol=Zero-Shot LLM2026.03 | 44.9 | 49.3 | 65 | 23.5 | 58.5 | |
| STAREvaluation Protocol=Traditional Reranker2026.03 | 41.2 | 46.6 | 63.4 | 22 | 53.3 | |
| TALLRecEvaluation Protocol=Post-Trained LLM, Base Model=Qwen2.5-3B-Instruct2026.03 | 39.9 | 45.9 | 62.6 | 21 | 50.7 | |
| Rank-GRPOEvaluation Protocol=Post-Trained LLM, Base Model=Qwen2.5-3B-Instruct2026.03 | 39.7 | 45.4 | 62.3 | 21.3 | 51.3 |