performance preservation
Maintaining the model's accuracy and efficiency while making alterations, such as model compression or architecture changes, to prevent degradation in performance.
- Frequency-Aware Token Reduction for Efficient Vision Transformer
- GC4NC: A Benchmark Framework for Graph Condensation on Node Classification with New Insights
- Mitigating Forgetting in LLM Fine-Tuning via Low-Perplexity Token Learning
- R-KV: Redundancy-aware KV Cache Compression for Reasoning Models
- Wisdom is Knowing What not to Say: Hallucination-Free LLMs Unlearning via Attention Shifting