work Selective KV Quantization Don't Drop It, Compress It - Reducing LLM KV cache memory usage by ~2x with minimal accuracy loss Rich Teacher Features for Efficient Single-Image Haze Removal Lightweight image dehazing using feature distillation and affinity modules Semi-Supervised Reinforcement Learning for Autonomous Agents Multi-modal RL framework with curiosity-driven exploration for adaptive policy learning in dynamic environments Super-Resolution Enhanced Face Recognition Improving face recognition accuracy on low-quality surveillance images using GANs fun
