Autonomous Agent Task Execution under Injection on ClawBench Office
0.7273Task Success Rate (TSR)OpenClaw
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| OpenClawGrafting Strategy=None2026.05 | 0.7273 | — | — | — | |
| OpenClawGrafting Strategy=EDIT S.2026.05 | 0.7273 | 0.1818 | 0.1818 | 0.1818 | |
| OpenClawGrafting Strategy=EDIT M.2026.05 | 0.7273 | 0.6364 | 0.6364 | 0.6364 | |
| ZeroClawGrafting Strategy=EDIT S.2026.05 | 0.7273 | 0 | 0 | 0 | |
| OpenClawGrafting Strategy=ADD S.2026.05 | 0.6364 | 1 | 1 | 1 | |
| OpenClawGrafting Strategy=ADD M.2026.05 | 0.6364 | 0.8182 | 0.8182 | 0.8182 | |
| OpenClawGrafting Strategy=ADD C.2026.05 | 0.6364 | 0.6364 | 0.6364 | 0.6364 | |
| OpenClawGrafting Strategy=EDIT C.2026.05 | 0.6364 | 0.9091 | 0.9091 | 0.9091 | |
| ZeroClawGrafting Strategy=ADD S.2026.05 | 0.6364 | 0.6364 | 0.6364 | 0.4545 | |
| ZeroClawGrafting Strategy=ADD M.2026.05 | 0.6364 | 0.7273 | 0.7273 | 0.7273 | |
| ZeroClawGrafting Strategy=EDIT M.2026.05 | 0.6364 | 0.7273 | 0.7273 | 0.7273 | |
| HermesGrafting Strategy=None2026.05 | 0.6364 | — | — | — | |
| HermesGrafting Strategy=EDIT C.2026.05 | 0.6364 | 0.7273 | 0.7273 | 0.7273 | |
| ZeroClawGrafting Strategy=EDIT C.2026.05 | 0.5455 | 1 | 0.9091 | 0.9091 | |
| HermesGrafting Strategy=ADD S.2026.05 | 0.5455 | 0.9091 | 0.9091 | 0.9091 | |
| HermesGrafting Strategy=ADD M.2026.05 | 0.5455 | 0.3636 | 0.3636 | 0.3636 | |
| HermesGrafting Strategy=ADD C.2026.05 | 0.5455 | 0.8182 | 0.7273 | 0.7273 | |
| HermesGrafting Strategy=EDIT S.2026.05 | 0.5455 | 0.3636 | 0.2727 | 0.1818 | |
| HermesGrafting Strategy=EDIT M.2026.05 | 0.5455 | 0.0909 | 0.0909 | 0.0909 | |
| ZeroClawGrafting Strategy=None2026.05 | 0.4545 | — | — | — | |
| ZeroClawGrafting Strategy=ADD C.2026.05 | 0.2727 | 0.9091 | 0.6364 | 0.6364 |