state-of-the-art techniques
The most advanced and effective algorithms or practices in a particular field at a given time. In AI, staying updated with state-of-the-art techniques is vital for achieving competitive performance in research and applications.
- AC-LoRA: (Almost) Training-Free Access Control Aware Multi-Modal LLMs
- Color Conditional Generation with Sliced Wasserstein Guidance
- Learning to Clean: Reinforcement Learning for Noisy Label Correction
- ReplaceMe: Network Simplification via Depth Pruning and Transformer Block Linearization
- SDPGO: Efficient Self-Distillation Training Meets Proximal Gradient Optimization
- T-SHIRT: Token-Selective Hierarchical Data Selection for Instruction Tuning