Question Answering on WikiHop
5.7Leakage ErrorSeedRG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SeedRGModel=Gemini 2.5 Flash2026.05 | 5.7 | 51.4 | |
| SeedRGModel=GPT-52026.05 | 11.4 | 45.7 | |
| SeedRGModel=Claude Sonnet 4.52026.05 | 12.9 | 65.7 | |
| Direct Generation (DG)Model=Gemini 2.5 Flash2026.05 | 35 | 38 | |
| Original BenchmarkModel=Claude Sonnet 4.52026.05 | 38.6 | 54.3 | |
| Direct Generation (DG)Model=GPT-52026.05 | 39 | 33 | |
| Original BenchmarkModel=Gemini 2.5 Flash2026.05 | 40 | 41.4 | |
| Direct Generation (DG)Model=Claude Sonnet 4.52026.05 | 42 | 46 | |
| Original BenchmarkModel=GPT-52026.05 | 62.9 | 20 |