Decoding Looped Transformers Better for (Almost) Free
Training-free LoopCD improves looped Transformer decoding and can cut recurrent compute by up to 48%.
2 stories, newest first · one card per story
Training-free LoopCD improves looped Transformer decoding and can cut recurrent compute by up to 48%.
LoopCD uses earlier recurrent states for contrastive decoding, boosting looped Transformers while cutting inference FLOPs.
Today 1.4K readers · 1.6K views · 308 reads · all time 110K views