EWAD and CPDP enable reliability-aware multi-teacher distillation that retains 71-122% of teacher ROUGE-L at 3.2x compression on Bangla datasets while showing logit-level KD as most reliable.
InPro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3967–3976
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Reliability Gated Multi-Teacher Distillation for Low Resource Abstractive Summarization
EWAD and CPDP enable reliability-aware multi-teacher distillation that retains 71-122% of teacher ROUGE-L at 3.2x compression on Bangla datasets while showing logit-level KD as most reliable.