Hallucination Prediction on MemoTrap
65AccuracyGAME-LoRA
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GAME-LoRAProtocol=Training-time2026.01 | 65 | — | — | — | — | |
| BaselineArchitecture=LoRA, Loss=Vanilla CE, Protocol=Training-time2026.01 | 64.2 | — | — | — | — | |
| ActDecProtocol=Inference-time2026.01 | 64.2 | — | — | — | — | |
| MEProtocol=Training-time2026.01 | 64.2 | — | — | — | — | |
| CADProtocol=Inference-time2026.01 | 64.1 | — | — | — | — | |
| DisagreementProtocol=Training-time2026.01 | 64.1 | — | — | — | — | |
| AlarmerModel=Llama-2-7b-chat, Protocol=Alarmer2025.02 | — | 35.8 | 31.2 | 24.7 | 19.6 | |
| AlarmerModel=Mistral 7b, Protocol=Alarmer2025.02 | — | 37.4 | 32.8 | 27.5 | 18.1 | |
| PromptModel=Llama-2-7b-chat, Protocol=Prompt2025.02 | — | 5.3 | 1.8 | 0 | 4.5 | |
| PromptModel=Mistral 7b, Protocol=Prompt2025.02 | — | 4.5 | 2.7 | 0 | 2.2 |