Autonomous Agent Task Execution under Injection on ClawBench (dev)
3Task Success Rate (TSR)OpenClaw
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| OpenClawGrafting Strategy=ADD C.2026.05 | 3 | 6 | 5 | 5 | |
| ZeroClawGrafting Strategy=EDIT S.2026.05 | 3 | 4 | 4 | 4 | |
| ZeroClawGrafting Strategy=EDIT M.2026.05 | 3 | 7 | 7 | 7 | |
| OpenClawGrafting Strategy=EDIT C.2026.05 | 2 | 10 | 9 | 9 | |
| ZeroClawGrafting Strategy=ADD S.2026.05 | 2 | 6 | 6 | 2 | |
| ZeroClawGrafting Strategy=ADD M.2026.05 | 2 | 6 | 6 | 6 | |
| ZeroClawGrafting Strategy=ADD C.2026.05 | 2 | 11 | 8 | 8 | |
| ZeroClawGrafting Strategy=EDIT C.2026.05 | 2 | 11 | 10 | 10 | |
| HermesGrafting Strategy=EDIT S.2026.05 | 2 | 6 | 5 | 3 | |
| OpenClawGrafting Strategy=None2026.05 | 1 | — | — | — | |
| OpenClawGrafting Strategy=ADD S.2026.05 | 1 | 9 | 9 | 9 | |
| OpenClawGrafting Strategy=ADD M.2026.05 | 1 | 9 | 9 | 9 | |
| OpenClawGrafting Strategy=EDIT S.2026.05 | 1 | 5 | 5 | 3 | |
| OpenClawGrafting Strategy=EDIT M.2026.05 | 1 | 8 | 8 | 8 | |
| ZeroClawGrafting Strategy=None2026.05 | 1 | — | — | — | |
| HermesGrafting Strategy=ADD S.2026.05 | 1 | 9 | 9 | 9 | |
| HermesGrafting Strategy=EDIT M.2026.05 | 1 | 5 | 5 | 5 | |
| HermesGrafting Strategy=None2026.05 | 0 | — | — | — | |
| HermesGrafting Strategy=ADD M.2026.05 | 0 | 6 | 6 | 6 | |
| HermesGrafting Strategy=ADD C.2026.05 | 0 | 10 | 10 | 10 | |
| HermesGrafting Strategy=EDIT C.2026.05 | 0 | 10 | 10 | 10 |