training-free
A characteristic of some AI systems where the performance is achieved without traditional training processes. These methods often rely on pre-trained models or leverage online learning techniques to adapt dynamically.
- Beyond Attention or Similarity: Maximizing Conditional Diversity for Token Pruning in MLLMs
- Color Conditional Generation with Sliced Wasserstein Guidance
- DyMU: Dynamic Merging and Virtual Unmerging for Efficient Variable-Length VLMs
- Enhancing CLIP Robustness via Cross-Modality Alignment
- GLSim: Detecting Object Hallucinations in LVLMs via Global-Local Similarity
- Hierarchical Optimization via LLM-Guided Objective Evolution for Mobility-on-Demand Systems
- Order-Level Attention Similarity Across Language Models: A Latent Commonality
- RPG360: Robust 360 Depth Estimation with Perspective Foundation Models and Graph Optimization
- ReplaceMe: Network Simplification via Depth Pruning and Transformer Block Linearization
- RetrievalAttention: Accelerating Long-Context LLM Inference via Vector Retrieval
- Token Perturbation Guidance for Diffusion Models
- Universal Cross-Tokenizer Distillation via Approximate Likelihood Matching