Fact recall on LongMemEval (500 questions)
97Fact RecallDiagnostic-conditional
Evaluation Results
| Method | Links | |
|---|---|---|
| Diagnostic-conditionalScale=14B2026.05 | 97 | |
| Best unconditionalScale=14B2026.05 | 96.8 | |
| Diagnostic-conditionalScale=8B2026.05 | 95.2 | |
| Best unconditionalScale=8B2026.05 | 93.2 | |
| IntactScale=14B2026.05 | 92.2 | |
| IntactScale=8B2026.05 | 90 | |
| Best unconditionalScale=4B2026.05 | 86.6 | |
| Diagnostic-conditionalScale=4B2026.05 | 86 | |
| IntactScale=4B2026.05 | 83.8 | |
| Diagnostic-conditionalScale=0.6B2026.05 | 54.8 | |
| Best unconditionalScale=0.6B2026.05 | 54.6 | |
| IntactScale=0.6B2026.05 | 54 |