Knowledge Work on GAIA 2 (held-out)
37Pass RateRHO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RHOHarness Architecture=Skills+Tools2026.06 | 37 | 0.08 | |
| Sleep-time ComputeHarness Architecture=Memory2026.06 | 32 | 0.03 | |
| Dynamic CheatsheetHarness Architecture=Skills2026.06 | 30 | 0.01 | |
| Vanilla CodexHarness Architecture=None2026.06 | 29 | — | |
| ReasoningBankHarness Architecture=Memory2026.06 | 28 | 0.01 |