E-commerce Review-based Question Answering on AmazonQA (test)
92.74BLEU-1Top-1 Using BLEU
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Top-1 Using BLEUMethod Category=Heuristic baselines2026.05 | 92.74 | 78.43 | 60.91 | 48.08 | 62.68 | |
| Top-1 using IRMethod Category=Heuristic baselines2026.05 | 89.49 | 74.8 | 56.76 | 43.52 | 61.48 | |
| SAGEtrained=true, Refinement Rounds=+1 round2026.05 | 82.76 | 68.91 | 52.74 | 39.68 | 58.83 | |
| Amazon User CommunityMethod Category=Human answers2026.05 | 80.88 | 68.86 | 54.36 | 42.01 | 62.18 | |
| Random SentenceMethod Category=Heuristic baselines2026.05 | 78.56 | 63.95 | 44.37 | 29.87 | 49.12 | |
| SAGEtrained=true2026.05 | 74.92 | 61.58 | 47.63 | 35.86 | 54.92 | |
| Expert (Spans)Method Category=Human answers2026.05 | 68.33 | 57.79 | 44.61 | 34.43 | 51.09 | |
| SAGE0-shot=true2026.05 | 61.84 | 49.36 | 37.82 | 28.41 | 46.73 | |
| Expert (Descriptive)Method Category=Human answers2026.05 | 53.67 | 46.56 | 37.81 | 30.76 | 53.31 | |
| R-NetMethod Category=Neural baseline2026.05 | 47.04 | 40.32 | 31.48 | 23.92 | 40.22 | |
| Top-1 Wilson ScoreMethod Category=Heuristic baselines2026.05 | 20.74 | 19.84 | 16.44 | 12.58 | 37.26 | |
| Top-1 HelpfulnessMethod Category=Heuristic baselines2026.05 | 20.66 | 19.78 | 16.39 | 12.54 | 37.01 |