zero-shot performance
Zero-shot performance describes a model's ability to handle tasks or classifications it has not been explicitly trained on. This implies the model can generalize knowledge gained from other tasks to solve new problems effectively.
- $\boldsymbol{\lambda}$-Orthogonality Regularization for Compatible Representation Learning
- AmorLIP: Efficient Language-Image Pretraining via Amortization
- Beyond Token Probes: Hallucination Detection via Activation Tensors with ACT-ViT
- Computation and Memory-Efficient Model Compression with Gradient Reweighting
- Improving Regret Approximation for Unsupervised Dynamic Environment Generation
- Multi-Scale Finetuning for Encoder-based Time Series Foundation Models
- Orient Anything V2: Unifying Orientation and Rotation Understanding
- PUO-Bench: A Panel Understanding and Operation Benchmark with A Privacy-Preserving Framework
- TS-RAG: Retrieval-Augmented Generation based Time Series Foundation Models are Stronger Zero-Shot Forecaster
- VITRIX-CLIPIN: Enhancing Fine-Grained Visual Understanding in CLIP via Instruction-Editing Data and Long Captions
- Zebra-Llama: Towards Extremely Efficient Hybrid Models