alinlab

StarFT: Robust Fine-tuning of Zero-shot Models via Spuriosity Alignment

Official implementation of Hierarchical Context Merging: Better Long Context Understanding for Pre-trained LLMs (ICLR 2024).

Modality-Agnostic Self-Supervised Learning with Meta-Learned Masked Auto-Encoder (NeurIPS 2023)

Unsupervised Meta-Learning via Few-shot Pseudo-supervised Contrastive Learning (ICLR 2023)

CSI: Novelty Detection via Contrastive Learning on Distributionally Shifted Instances (NeurIPS 2020)

MASKER: Masked Keyword Regularization for Reliable Text Classification (AAAI 2021)

IFSeg: Image-free Semantic Segmentation via Vision-Language Model (CVPR 2023)

S-CLIP: Semi-supervised Vision-Language Pre-training using Few Specialist Captions (NeurIPS 2023)

Bias-to-Text: Debiasing Unknown Visual Biases through Language Interpretation (CVPR 2024)