Long-context Modeling, Reinforcement-Learning, Multi-modality
Flux Attention: Context-Aware Hybrid Attention for Efficient LLMs Inference