2026. Multi-level Constraint-Based Two-Stage Few-Shot Knowledge Distillation for Vision-Language Models. Informatica. 50, 13 (Jun. 2026). DOI:https://doi.org/10.31449/inf.v50i13.13666.