Agentic Trajectory Question Answering on Benchmarks 12 task columns
3.1Average RankAutoMEM
Evaluation Results
| Method | Links | |
|---|---|---|
| AutoMEMvariant=full loop2026.06 | 3.1 | |
| AutoMEMvariant=PlugMem graph2026.06 | 3.21 | |
| AutoMEMvariant=single-pass2026.06 | 3.38 | |
| DCI-Lite2026.06 | 4.1 | |
| Long context2026.06 | 4.93 |