Interactive Instruction Following on ALFWorld OOD
92Success RateTask-Perturbed NLL Optimization
Evaluation Results
| Method | Links | |
|---|---|---|
| Task-Perturbed NLL OptimizationBackbone=Qwen3-8B2026.06 | 92 | |
| ProcMEM2026.02 | 90.9 | |
| AWM2026.02 | 90 | |
| GRPOBackbone=Qwen3-8B2026.06 | 86 | |
| G-Memory2026.02 | 81.2 | |
| Expel2026.02 | 74 | |
| ReAct2026.02 | 64 | |
| A-MEM2026.02 | 64 | |
| CoT2026.02 | 62 | |
| Task-Perturbed NLL OptimizationBackbone=Qwen3-4B2026.06 | 62 | |
| GRPOBackbone=Qwen3-4B2026.06 | 44 | |
| RAG2026.02 | 40.2 | |
| State2026.02 | 26.2 |