Grid-world Navigation on FrozenLake reward reversal Hidden drift
100ScoreVanilla + GLOVE
Evaluation Results
| Method | Links | |
|---|---|---|
| Vanilla + GLOVEBackbone=Llama3.3-70B, Drift Type=Implicit, Stage=Hidden drift2026.01 | 100 | |
| VoyagerBackbone=Llama3.3-70B, Drift Type=Implicit, Stage=Hidden drift2026.01 | 100 | |
| Voyager + GLOVEBackbone=Llama3.3-70B, Drift Type=Implicit, Stage=Hidden drift2026.01 | 100 | |
| Generative AgentBackbone=Llama3.3-70B, Drift Type=Implicit, Stage=Hidden drift2026.01 | 100 | |
| Generative Agent + GLOVEBackbone=Llama3.3-70B, Drift Type=Implicit, Stage=Hidden drift2026.01 | 100 | |
| Vanilla +GLOVELLM=Grok-3, Augmentation=GLOVE2026.01 | 100 | |
| Voyager +GLOVELLM=Grok-3, Augmentation=GLOVE2026.01 | 100 | |
| Voyager + GLOVEBackbone=Llama3.1-8B2026.01 | 100 | |
| Vanilla + GLOVELLM Backbone=GPT-4o, Base Agent=Vanilla2026.01 | 97.5 | |
| MemoryBank + GLOVELLM Backbone=GPT-4o, Base Agent=MemoryBank2026.01 | 97.5 | |
| Voyager + GLOVELLM Backbone=GPT-4o, Base Agent=Voyager2026.01 | 97.5 | |
| MemoryBank +GLOVELLM=Grok-3, Augmentation=GLOVE2026.01 | 97.5 | |
| Generative Agent +GLOVELLM=Grok-3, Augmentation=GLOVE2026.01 | 97.5 | |
| VanillaBackbone=Llama3.3-70B, Drift Type=Implicit, Stage=Hidden drift2026.01 | 95 | |
| Vanilla + GLOVEBackbone=Llama3.1-8B2026.01 | 95 | |
| Generative Agent + GLOVELLM Backbone=GPT-4o, Base Agent=Generative Agent2026.01 | 92.5 | |
| MemoryBankBackbone=Llama3.3-70B, Drift Type=Implicit, Stage=Hidden drift2026.01 | 90 | |
| MemoryBank + GLOVEBackbone=Llama3.3-70B, Drift Type=Implicit, Stage=Hidden drift2026.01 | 90 | |
| VanillaLLM=Grok-32026.01 | 90 | |
| Generative Agent + GLOVEBackbone=Llama3.1-8B2026.01 | 90 | |
| MemoryBankLLM=Grok-32026.01 | 82.5 | |
| MemoryBankBackbone=Llama3.1-8B2026.01 | 65 | |
| MemoryBank + GLOVEBackbone=Llama3.1-8B2026.01 | 65 | |
| Generative AgentBackbone=Llama3.1-8B2026.01 | 60 | |
| MemoryBankLLM Backbone=GPT-4o2026.01 | 57.5 | |
| VanillaLLM Backbone=GPT-4o2026.01 | 50 | |
| VoyagerLLM Backbone=GPT-4o2026.01 | 50 | |
| Generative AgentLLM Backbone=GPT-4o2026.01 | 50 | |
| VoyagerLLM=Grok-32026.01 | 50 | |
| Generative AgentLLM=Grok-32026.01 | 50 | |
| VanillaBackbone=Qwen3-30B, Agent Architecture=Vanilla, GLOVE Augmentation=false2026.01 | 50 | |
| Vanilla + GLOVEBackbone=Qwen3-30B, Agent Architecture=Vanilla, GLOVE Augmentation=true2026.01 | 50 | |
| VoyagerBackbone=Qwen3-30B, Agent Architecture=Voyager, GLOVE Augmentation=false2026.01 | 50 | |
| Voyager + GLOVEBackbone=Qwen3-30B, Agent Architecture=Voyager, GLOVE Augmentation=true2026.01 | 50 | |
| Generative AgentBackbone=Qwen3-30B, Agent Architecture=Generative Agent, GLOVE Augmentation=false2026.01 | 50 | |
| Generative Agent + GLOVEBackbone=Qwen3-30B, Agent Architecture=Generative Agent, GLOVE Augmentation=true2026.01 | 50 | |
| VanillaBackbone=Llama3.1-8B2026.01 | 50 | |
| MemoryBank + GLOVEBackbone=Qwen3-30B, Agent Architecture=MemoryBank, GLOVE Augmentation=true2026.01 | 40 | |
| MemoryBankBackbone=Qwen3-30B, Agent Architecture=MemoryBank, GLOVE Augmentation=false2026.01 | 37.5 | |
| VoyagerBackbone=Llama3.1-8B2026.01 | 35 | |
| No Memory (Plain)Backbone=Llama3.3-70B, Drift Type=Implicit, Stage=Hidden drift2026.01 | 0 | |
| No Memory (Plain)LLM Backbone=GPT-4o2026.01 | 0 | |
| No Memory (Plain)LLM=Grok-32026.01 | 0 | |
| No Memory (Plain)Backbone=Qwen3-30B, Agent Architecture=No Memory, GLOVE Augmentation=false2026.01 | 0 | |
| No Memory (Plain)Backbone=Llama3.1-8B2026.01 | 0 |