Iterative Multi-hop Search on HotpotQA (cross-episode setup)
41.1Success RateADAMEM
Evaluation Results
| Method | Links | |
|---|---|---|
| ADAMEMBackbone=Qwen3-4B-Instruct-2507, Long Term=Traj., Short Term=Strat.2026.06 | 41.1 | |
| ReasoningBankBackbone=Qwen3-4B-Instruct-2507, Long Term=Strat., Short Term=Strat.2026.06 | 40.5 | |
| SynapseBackbone=Qwen3-4B-Instruct-2507, Long Term=Traj., Short Term=Traj.2026.06 | 40.4 | |
| No MemoryBackbone=Qwen3-4B-Instruct-25072026.06 | 39.7 |