General Assistant Setting on GAIA (Acc/Traj)
72.7AccuracyClaude Code
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Claude CodeState=✗, Ctrl=✓, Recov=∼2026.06 | 72.7 | 45,000 | |
| VISTA (full)State=✓, Ctrl=✓, Recov=✓2026.06 | 72.7 | 35,000 | |
| Active Ctx. CompressionState=✗, Ctrl=✓, Recov=✗2026.06 | 66.7 | 38,000 | |
| Auto-Archive + RecoverState=✗, Ctrl=✗, Recov=✓2026.06 | 66.7 | 28,000 | |
| VISTA w/o dashboardState=✗, Ctrl=✓, Recov=✓2026.06 | 66.7 | 60,000 | |
| VISTA w/o recoveryState=✓, Ctrl=✓, Recov=✗2026.06 | 66.7 | 32,000 | |
| Tool-result ClearingState=✗, Ctrl=✗, Recov=✗2026.06 | 63.6 | 29,000 | |
| SLIM (summary)State=✗, Ctrl=∼, Recov=✗2026.06 | 63.6 | 43,000 | |
| Context-FoldingState=✗, Ctrl=✓, Recov=✗2026.06 | 63.6 | 45,000 | |
| ReActState=✗, Ctrl=✗, Recov=✗2026.06 | 60.6 | 23,000 | |
| Stale-obs. MaskingState=✗, Ctrl=✗, Recov=✗2026.06 | 60.6 | 20,000 | |
| Skeleton CompressionState=✗, Ctrl=✗, Recov=✗2026.06 | 57.6 | 18,000 |