Generalization to agglutinative or tonal languages

Determine whether the reliability-gated multi-teacher distillation framework for low-resource abstractive summarization—specifically the Entropy-Weighted Agreement-Aware Distillation (EWAD) and Capacity-Proportional Divergence Preservation (CPDP) components—generalizes to highly agglutinative or tonal languages that were not included in the reported cross-lingual experiments.

Background

The paper evaluates cross-lingual pseudo-label knowledge distillation across ten languages and multiple scripts, demonstrating that the proposed reliability-aware distillation pipeline can retain a substantial fraction of teacher performance at 3.2× compression.

However, the evaluated set does not cover certain linguistic typologies such as highly agglutinative or tonal languages, leaving open whether the EWAD and CPDP framework transfers effectively to these language families.

References

Second, although we evaluate cross-lingual generalization across ten languages and multiple scripts, the coverage does not include certain linguistic typologies such as highly agglutinative or tonal languages. The generalization of the proposed framework to such languages remains an open question.

— Reliability Gated Multi-Teacher Distillation for Low Resource Abstractive Summarization  (2604.03192 - Sumit et al., 3 Apr 2026) in Limitations, paragraph 2