Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
2
0
4
11
Harnessing earlier recurrent states as implicit weak supervision, a training-free contrastive decoding method boosts Looped Transformer performance by up to 11.45% and enables 22.5%-48.2% FLOP reduction.
Self-distillation, using only an LLM's own raw outputs, can boost its code generation performance by over 10 points on pass@1, rivaling more complex methods.