AI Briefing#encoder-decoder
The latest AI and developer news about #encoder-decoder, with the original source and a short summary.
The latest AI and developer news about #encoder-decoder, with the original source and a short summary.
Applying the model grafting technique, which converts existing models into a causal encoder-decoder architecture, increased Qwen3.5-4B's prompt processing speed by up to 3.7x.
LG AI Research achieved the #1 ranking on the Korean machine reading comprehension benchmark KorQuAD 2.0 using EXAONE LM 1.0.
A new high-performance Reranker model with a novel architecture that separates query and passage computation to improve computational efficiency has been released.
HeadsUp reconstructs 3D Gaussian heads without test-time optimization using data at a scale of over 10,000 people.
Dynamic batching boosted throughput by 3.3x when training NLLB-200 600M on an RTX 5090.
The Ettin Suite has been released, a family of SoTA-level encoder and decoder models trained with the same data and training recipe.
This introduces a Warm-starting technique that leverages pretrained BERT or GPT2 to reduce the training cost of Encoder-Decoder models.