Question Answering on LongMemEval M-tier (475 Sessions)
78.2KU ScoreRaw RAG
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Raw RAGk=102026.05 | 78.2 | 57.1 | 53.3 | 96.4 | 92.9 | 63.2 | 71.2 | 67.2 | |
| Dedup-adaptiveTokens=200K, top-k=352026.05 | 74.4 | 58.3 | 53.3 | 91.1 | 85.7 | 66.2 | 70.1 | 66 | |
| Dedup-adaptiveTokens=115K, top-k=352026.05 | 69.2 | 57.9 | 40 | 83.9 | 82.9 | 60.2 | 65.6 | 61.4 | |
| Dedup-adaptiveTokens=50K, top-k=352026.05 | 56.4 | 39.1 | 43.3 | 60.7 | 50 | 51.1 | 49.2 | 44.8 | |
| Dedup-adaptiveTokens=25K, top-k=352026.05 | 48.7 | 24.8 | 33.3 | 41.1 | 34.3 | 49.6 | 38.8 | 34.6 |