Sequential task management and state maintenance on Lifelong AgentBench
100AccuracyGA
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GAModel=Claude Sonnet 4.62026.04 | 100 | 222 | 20 | 241 | 4.15 | |
| GAModel=Claude Sonnet 4.62026.06 | 100 | 222 | 20 | 241 | 4.15 | |
| BA-IncModel=deepseek-v4-flash2026.06 | 100 | 78 | 6 | 84 | 23.85 | |
| GAModel=deepseek-v4-pro2026.06 | 100 | 385 | 34 | 418 | 47.82 | |
| BA-FullModel=deepseek-v4-pro2026.06 | 100 | 405 | 35 | 439 | 45.53 | |
| BA-IncModel=deepseek-v4-pro2026.06 | 100 | 0 | 0 | 0 | — | |
| GAModel=Minimax M2.72026.04 | 90 | 400 | 23 | 423 | 2.12 | |
| GAModel=deepseek-v4-flash2026.06 | 90 | 649 | 42 | 690 | 26.07 | |
| BA-FullModel=deepseek-v4-flash2026.06 | 85 | 634 | 40 | 674 | 25.23 | |
| Claude CodeModel=Claude Sonnet 4.62026.04 | 75 | 800 | 14 | 814 | 0.92 | |
| Claude CodeModel=Claude Sonnet 4.62026.06 | 75 | 800 | 14 | 814 | 0.92 | |
| OpenClawModel=Claude Sonnet 4.62026.04 | 70 | 1.43 | 21 | 1.45 | 0.48 | |
| OpenClawModel=Minimax M2.72026.04 | 70 | 1.2 | 17 | 1.22 | 0.57 | |
| OpenClawModel=Claude Sonnet 4.62026.06 | 70 | 1.43 | 21 | 1.45 | 0.48 |