Summarization on XLSUM ko
0Response-level Language ConfusionQwen3 4B-I
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Qwen3 4B-I2026.06 | 0 | 12.34 | |
| Qwen3 4B-IPrompting strategy=Strict Prompt2026.06 | 0 | 11.78 | |
| Qwen3 4B-I + LATB2026.06 | 0 | 11.51 | |
| Qwen3 4B-I + Adaptive LATB2026.06 | 0 | 11.55 |