DuDi: Dual-Signal Distillation Framework Improves Multilingual Capabilities of Small Language Models
Researchers introduced DuDi, a multilingual distillation framework designed to improve the performance of small language models on Southeast Asian languages. The method combines sequence-level and token-level learning signals with a cross-lingual verbalizer to enhance knowledge transfer from larger teacher models. This addresses a significant limitation in scaling efficient language models while maintaining multilingual capabilities.
DuDi is a dual-signal multilingual distillation framework that tackles the problem of degraded multilingual performance in small language models (SLMs) at sub-billion parameter scales, particularly for Southeast Asian languages. The approach combines an online sequence-level signal with both off-policy and on-policy token-level signals, and introduces a cross-lingual verbalizer to refine teacher feedback and improve transferability in multilingual contexts. Experiments conducted on SEA-HELM across multiple model families, scales, and teacher-student configurations demonstrate that DuDi consistently outperforms competitive distillation baselines. Ablation studies and analyses confirm that the three components—sequence-level optimization, token-level supervision, and cross-lingual verbalization—provide complementary and transferable learning signals. The work addresses a practical challenge in deploying efficient language models across linguistically diverse regions.
What different sources said
- arXiv cs.CLCenter
DuDi: Dual-Signal Distillation with Cross-Lingual Verbalizer
Related
Genetic Drift, Not Selection, Drives Rapid Feather Color Evolution in Island Bird Radiation
A new study of an island bird radiation found that rapid evolution of feather coloration is driven primarily by genetic drift in small populations rather than sexual or ecological selection. The research integrated whole-genome data with detailed plumage measurements across complete species sampling to test whether signaling trait evolution correlates with speciation rates. The findings suggest that neutral demographic processes play a central role in generating phenotypic diversity during island radiations, challenging assumptions about the mechanisms driving rapid evolution.
New AI Model Improves Prediction of Therapeutic Peptide Function from Protein Sequences
Researchers developed a lightweight CNN classifier that predicts whether peptide sequences have therapeutic properties, trained on a database of 54,655 peptides across 48 functional categories. The model uses a novel negative sampling strategy to reduce false positive rates from over 60% in previous approaches to 2.1%. This advancement could accelerate drug discovery by enabling faster computational screening of peptide candidates before expensive experimental testing.
Study Shows Different Metabolic Stress Models Produce Distinct Effects on Human Neuronal Networks
Researchers tested three common in vitro metabolic stress models on human-derived neuronal networks and found each produced different patterns of neuronal activity and cell damage. The models tested were hypoxia alone, oxygen-glucose deprivation (OGD), and hypoxia combined with glutamate exposure. The findings suggest that choice of experimental model significantly affects results and that combining electrophysiological and structural analyses is important for accurately assessing metabolic stress in stroke research.