Best pretrained models for zero-shot classification?

Natural Language Processing

nlp,vision transformers
J

Jodi Blankenship

schedule5 months ago

Are there language or vision models that perform well on unseen classes with minimal fine-tuning?

B

Brandi Mills

schedule4 months ago

Try CLIP for text-to-image zero-shot tasks. It's surprisingly versatile.

K

Kelsey Adams

schedule4 months ago

Multilingual zero-shot classification is still a challenge—especially for low-resource languages.

J

Joshua Clay

schedule4 months ago

I use SBERT + cosine similarity for lightweight semantic matching.

Reply