Text Summarization Hallucination Evaluation on MultiNews
19AccuracyMEAP
Evaluation results
| Method | Links | |
|---|---|---|
| MEAPTraining Objective=Mask-Enhanced Autoregressive Prediction, LLM Judge=Deepseek-V32025.02 | 19 | |
| NTPTraining Objective=Next Token Prediction, LLM Judge=Deepseek-V32025.02 | 17 | |
| MEAPTraining Objective=Mask-Enhanced Autoregressive Prediction, LLM Judge=Qwen-Plus2025.02 | 14 | |
| MEAPTraining Objective=Mask-Enhanced Autoregressive Prediction, LLM Judge=GPT-4o2025.02 | 13 | |
| NTPTraining Objective=Next Token Prediction, LLM Judge=Qwen-Plus2025.02 | 11 | |
| NTPTraining Objective=Next Token Prediction, LLM Judge=GPT-4o2025.02 | 10 |