Summarization on TruthfulQA Generalization
14.77ROUGE-LChain of Summaries (CoS)
Evaluation Results
| Method | Links | |
|---|---|---|
| Chain of Summaries (CoS)Backbone=Llama3.2:3B2025.11 | 14.77 | |
| Chain of Summaries (CoS)Backbone=GPT-4o-mini2025.11 | 12.69 | |
| Chain of Summaries (CoS)Backbone=Qwen2.5:7B2025.11 | 10.71 | |
| Zero-shot summaryBackbone=GPT-4o-mini2025.11 | 10.67 | |
| Original ContentBackbone=GPT-4o-mini2025.11 | 10 | |
| BRIOBackbone=GPT-4o-mini2025.11 | 9.25 | |
| Chain of DensityBackbone=GPT-4o-mini2025.11 | 9.07 | |
| Zero-shot summaryBackbone=Llama3.2:3B2025.11 | 8.89 | |
| PEGASUSBackbone=GPT-4o-mini2025.11 | 8.28 | |
| Zero-shot summaryBackbone=Qwen2.5:7B2025.11 | 7.35 | |
| Original ContentBackbone=Qwen2.5:7B2025.11 | 6.25 | |
| PEGASUSBackbone=Qwen2.5:7B2025.11 | 5.92 | |
| BRIOBackbone=Llama3.2:3B2025.11 | 5.74 | |
| BRIOBackbone=Qwen2.5:7B2025.11 | 5.74 | |
| Chain of DensityBackbone=Qwen2.5:7B2025.11 | 5.13 | |
| Chain of DensityBackbone=Llama3.2:3B2025.11 | 4.63 | |
| Original ContentBackbone=Llama3.2:3B2025.11 | 4.52 | |
| PEGASUSBackbone=Llama3.2:3B2025.11 | 3.29 |