High Quality Decoder
High-quality decoder research focuses on improving the accuracy and efficiency of decoding processes across diverse applications, from image and speech processing to natural language understanding and error correction. Current efforts concentrate on developing novel decoder architectures, such as those incorporating attention mechanisms, low-rank adaptations, and efficient pruning techniques, often within larger encoder-decoder frameworks or leveraging transfer learning from pre-trained models. These advancements are significant because they enable improved performance in various fields, ranging from medical image analysis and machine translation to resource-constrained devices and communication systems.
Papers
Sketch and shift: a robust decoder for compressive clustering
Ayoub Belhadji, Rémi Gribonval
On the compression of shallow non-causal ASR models using knowledge distillation and tied-and-reduced decoder for low-latency on-device speech recognition
Nagaraj Adiga, Jinhwan Park, Chintigari Shiva Kumar, Shatrughan Singh, Kyungmin Lee, Chanwoo Kim, Dhananjaya Gowda