Short-Form Question Answering on SAGE Web Search
63.3Accuracy (Com. Sci.)BM25
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| BM25k=10, Avg. Searches=6.42, Avg. Refs=40.02026.02 | 63.3 | 88.8 | 84.4 | 88.4 | 81.2 | |
| GPT-5Avg. Searches=8.78, Avg. Refs=6.542026.02 | 57.3 | 78.7 | 79.1 | 71.7 | 71.7 | |
| BM25k=5, Avg. Searches=7.54, Avg. Refs=22.42026.02 | 56 | 79.9 | 82 | 85.3 | 75.8 | |
| gte-Qwenk=10, Avg. Searches=5.88, Avg. Refs=33.22026.02 | 44.4 | 69.7 | 73.3 | 64.4 | 63 | |
| GPT-5-miniAvg. Searches=8.15, Avg. Refs=4.692026.02 | 40 | 72 | 70.7 | 66.4 | 62.3 | |
| DR TuluAvg. Searches=7.35, Avg. Refs=37.322026.02 | 36 | 58 | 49.3 | 44.7 | 42 | |
| gte-Qwenk=5, Avg. Searches=4.82, Avg. Refs=14.22026.02 | 32.9 | 52.1 | 62.6 | 40 | 46.9 | |
| GPT-5-nanoAvg. Searches=8.92, Avg. Refs=7.062026.02 | 30.7 | 46.7 | 62 | 44.3 | 45.9 | |
| Gemini-2.5-flashAvg. Searches=15.64, Avg. Refs=3.792026.02 | 30.3 | 43.7 | 41.6 | 36.4 | 38 | |
| ReasonIRk=10, Avg. Searches=7.51, Avg. Refs=35.82026.02 | 28 | 57 | 61.3 | 50.7 | 49.3 | |
| Gemini-2.5-proAvg. Searches=14.02, Avg. Refs=1.432026.02 | 27.7 | 47.8 | 40.3 | 37.7 | 38.5 | |
| ReasonIRk=5, Avg. Searches=8.78, Avg. Refs=17.42026.02 | 25.9 | 42.2 | 48.2 | 38.4 | 38.7 |