Continual Learning on KQA → Math → Code → MQA Task-level stream
51.67AA ScoreMasFACT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MasFACTReplay-based=false2026.05 | 51.67 | 2.07 | |
| MasFACTReplay-based=false2026.05 | 51.67 | 2.07 | |
| MasRouterReplay-based=true2026.05 | 49.55 | 5.02 | |
| MasRouterReplay-based=true2026.05 | 49.55 | 5.02 | |
| ARGDesignerReplay-based=true2026.05 | 49.13 | 4.83 | |
| ARGDesignerReplay-based=true2026.05 | 49.13 | 4.83 | |
| MasRouterReplay-based=false2026.05 | 45.5 | 7.48 | |
| MasRouterReplay-based=false2026.05 | 45.5 | 7.48 | |
| GDesignerReplay-based=true2026.05 | 44.21 | 5.15 | |
| GDesignerReplay-based=true2026.05 | 44.21 | 5.15 | |
| ARGDesignerReplay-based=false2026.05 | 42.18 | 6.14 | |
| ARGDesignerReplay-based=false2026.05 | 42.18 | 6.14 | |
| GDesignerReplay-based=false2026.05 | 41.81 | 6.95 | |
| GDesignerReplay-based=false2026.05 | 41.81 | 6.95 | |
| AgentDropoutReplay-based=true2026.05 | 40.61 | 7.48 | |
| AgentDropoutReplay-based=true2026.05 | 40.61 | 7.48 | |
| AFLOWReplay-based=true2026.05 | 40.39 | 4.12 | |
| AFLOWReplay-based=true2026.05 | 40.39 | 4.12 | |
| AFLOWReplay-based=false2026.05 | 37.76 | 6.45 | |
| AFLOWReplay-based=false2026.05 | 37.76 | 6.45 | |
| SC(CoT)-ICLReplay-based=false2026.05 | 37.47 | 5.63 | |
| SC(CoT)-ICLReplay-based=false2026.05 | 37.47 | 5.63 | |
| AgentDropoutReplay-based=false2026.05 | 36.72 | 8.12 | |
| AgentDropoutReplay-based=false2026.05 | 36.72 | 8.12 | |
| CompleteReplay-based=false2026.05 | 36.06 | — | |
| Complete GraphReplay-based=false2026.05 | 36.06 | — | |
| CoT-ICLReplay-based=false2026.05 | 34.9 | 5.64 | |
| CoT-ICLReplay-based=false2026.05 | 34.9 | 5.64 | |
| TreeReplay-based=false2026.05 | 34.17 | — | |
| TreeReplay-based=false2026.05 | 34.17 | — | |
| ChainReplay-based=false2026.05 | 32.24 | — | |
| ChainReplay-based=false2026.05 | 32.24 | — | |
| VanillaReplay-based=false2026.05 | 30.84 | — | |
| VanillaReplay-based=false2026.05 | 30.84 | — |