Embodied AI task execution on ALFWorld (held-out)
97.9SRMemRL
Evaluation Results
| Method | Links | |
|---|---|---|
| MemRLType=Q-value episodic memory, Base Model=GPT-5-mini, Memory=Frozen2026.03 | 97.9 | |
| RAGType=Semantic retrieval, Base Model=GPT-5-mini, Memory=Frozen2026.03 | 95 | |
| Self-RAGType=Critique-filtered retrieval, Base Model=GPT-5-mini, Memory=Frozen2026.03 | 95 | |
| Mem0Type=Entity-centric memory, Base Model=GPT-5-mini, Memory=Frozen2026.03 | 95 | |
| MemPType=Procedural distillation, Base Model=GPT-5-mini, Memory=Frozen2026.03 | 92.1 | |
| No MemoryType=Single-shot LLM, Base Model=GPT-5-mini, Memory=Frozen2026.03 | 83.6 |