ReasonCLIP Models
Collection
Visually Grounded Commonsense Reasoning Supervision for CLIP (ECCV 2026) • 25 items • Updated • 1
# Load model directly
from transformers import AutoProcessor, AutoModelForZeroShotImageClassification
processor = AutoProcessor.from_pretrained("RISys-Lab/ReasonSigLIP-So14-384-S2")
model = AutoModelForZeroShotImageClassification.from_pretrained("RISys-Lab/ReasonSigLIP-So14-384-S2")Quick Links
Model Details
- Model: ReasonSigLIP-So14-384-S2
- Base model: google/siglip-so400m-patch14-384
- Architecture: SigLIP ViT-So400M/14
- Image resolution: 384
- Training stage: Stage 2
- Training data: ReasonLite-42M and ReasonPro-16M
Method
Resources
- GitHub: RISys-Lab/ReasonCLIP
- Paper: arXiv:2606.26794
- Playground: ReasonCLIP Playground
Usage
from transformers import AutoModel, AutoProcessor
model_id = "RISys-Lab/ReasonSigLIP-So14-384-S2"
model = AutoModel.from_pretrained(model_id)
processor = AutoProcessor.from_pretrained(model_id)
For the full checkpoint list, see the ReasonCLIP model card.
- Downloads last month
- 75
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("zero-shot-image-classification", model="RISys-Lab/ReasonSigLIP-So14-384-S2") pipe( "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/hub/parrots.png", candidate_labels=["animals", "humans", "landscape"], )