Question Answering on 2WikiMultiHopQA (held-out)
77.4F1 ScoreRASER-3
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RASER-3Reader=GPT-OSS-120B, N=5002026.06 | 77.4 | 1.6 | |
| RASER-2Reader=GPT-OSS-120B, N=5002026.06 | 76.3 | 1.3 | |
| ONE-SHOT RAGReader=GPT-OSS-120B, N=5002026.06 | 75.4 | 1.1 | |
| RASER-3Reader=Gemma-3-31B, N=3002026.06 | 70 | 1.6 | |
| RASER-2Reader=Gemma-3-31B, N=3002026.06 | 65.6 | 1.2 | |
| ONE-SHOT RAGReader=Gemma-3-31B, N=3002026.06 | 63 | 962 | |
| RASER-3Reader=Llama-3.1-8B, N=3002026.06 | 48.7 | 2.2 | |
| RASER-2Reader=Mistral-S-119B, N=3002026.06 | 48.5 | 1.5 | |
| RASER-3Reader=Phi-4-mini, N=3002026.06 | 47.4 | 2.3 | |
| RASER-3Reader=Llama-3-8B, N=3002026.06 | 47.2 | 2.1 | |
| RASER-3Reader=Mistral-S-119B, N=3002026.06 | 45.1 | 979 | |
| ONE-SHOT RAGReader=Mistral-S-119B, N=3002026.06 | 45 | 970 | |
| RASER-2Reader=Phi-4-mini, N=3002026.06 | 40.2 | 1.1 | |
| ONE-SHOT RAGReader=Phi-4-mini, N=3002026.06 | 39.1 | 904 | |
| RASER-2Reader=Llama-3-8B, N=3002026.06 | 38 | 1.2 | |
| ONE-SHOT RAGReader=Llama-3-8B, N=3002026.06 | 36.7 | 932 | |
| RASER-2Reader=Llama-3.1-8B, N=3002026.06 | 35.5 | 1.4 | |
| ONE-SHOT RAGReader=Llama-3.1-8B, N=3002026.06 | 30.7 | 954 |