CVAIJun 22

The Professor: Multi-Teacher Unsupervised Prompt Distillation for Vision-Language Models

arXiv:2606.238974.7
Predicted impact top 82% in CV · last 90 daysOriginality Synthesis-oriented
AI Analysis

For practitioners of prompt distillation, this work shows that multi-teacher ensembling provides modest gains that are dataset-dependent, with the most benefit under domain shift.

The Professor extends PromptKD by distilling from a two-teacher ensemble (domain-finetuned and zero-shot VLMs), achieving a +1.77 HM improvement on four base-to-novel datasets, with the largest gain (+5.78 HM) on domain-shifted EuroSAT.

Prompt distillation compresses large vision-language models (VLMs) such as CLIP into lightweight student models by matching teacher predictions on unlabeled domain images. PromptKD (CVPR 2024) established this paradigm with a single PromptSRC-finetuned ViT-L/14 teacher and a ViT-B/16 student. We propose TheProfessor, a multi-teacher extension that distills from a fixed two-teacher ensemble: a domain-finetuned PromptSRC ViT-L/14 teacher and a zero-shot EVA-CLIP-L/14 teacher whose logits are pre-computed per dataset. We evaluate single-teacher PromptKD, equal-probability ensembling, and confidence-weighted ensembling on four base-to-novel datasets: Caltech-101, DTD, UCF101, and EuroSAT. In a 12-run single-seed sweep, confidence-weighted ensembling improves average HM from 87.52 to 89.28 (+1.77 points), while equal averaging improves average HM to 88.88 (+1.37 points). Gains are dataset dependent: they are negligible on Caltech-101 (+0.16 HM for confidence weighting), modest on UCF101 (+0.62), and largest on domain-shifted EuroSAT (+5.78). These results update our earlier Caltech-only analysis and show that multi-teacher prompt distillation is most useful when the second teacher contributes complementary supervision under domain shift.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes