Hallucination Prediction on Overshadowing
65Accuracy (Time)CoDA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CoDABackbone=Llama-2-7b-chat2025.02 | 65 | 46.8 | — | |
| CoDABackbone=Mistral-7b2025.02 | 61.2 | 49.5 | — | |
| DolaBackbone=Llama-2-7b-chat2025.02 | 53.6 | 31.8 | — | |
| DolaBackbone=Mistral-7b2025.02 | 51 | 34.6 | — | |
| AlarmerModel=Mistral 7b, Protocol=Alarmer2025.02 | 42.5 | 51.6 | — | |
| SRBackbone=Llama-2-7b-chat2025.02 | 42.5 | 23.8 | — | |
| SRBackbone=Mistral-7b2025.02 | 42.4 | 24.9 | — | |
| GreedyBackbone=Llama-2-7b-chat2025.02 | 41.4 | 20.8 | — | |
| AlarmerModel=Llama-2-7b-chat, Protocol=Alarmer2025.02 | 40.4 | 53.3 | — | |
| CoTBackbone=Llama-2-7b-chat2025.02 | 40.4 | — | — | |
| USCBackbone=Llama-2-7b-chat2025.02 | 40.2 | 16.4 | — | |
| GreedyBackbone=Mistral-7b2025.02 | 39.5 | 21.6 | — | |
| USCBackbone=Mistral-7b2025.02 | 37.9 | 20.8 | — | |
| CoTBackbone=Mistral-7b2025.02 | 37 | — | — | |
| PromptModel=Mistral 7b, Protocol=Prompt2025.02 | 2.2 | — | — | |
| PromptModel=Llama-2-7b-chat, Protocol=Prompt2025.02 | 0.6 | — | — |