Multilingual Summarization on XL-SUM fr
11Token-level Language Confusion (%)LATB
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| LATBBackbone=Llama3 8B-I2026.06 | 11 | 0.35 | 0.2 | |
| Adaptive LATBBackbone=Llama3 8B-I2026.06 | 18 | 0.25 | 0.3 | |
| Llama3 8B-IPrompting Strategy=Standard2026.06 | 24 | 47.14 | 40.6 | |
| Llama3 8B-IPrompting Strategy=Strict Prompt2026.06 | 26 | 0.39 | 0.2 | |
| Suzume 8B-Multilingual2026.06 | 31 | 0.37 | 0.3 |