Search papers, labs, and topics across Lattice.
Lamarr Institute
4
0
5
LoopMTP boosts reasoning accuracy by up to 8.1% by effectively guiding looped transformer iterations with multi-token prediction.
English LLMs dominate math reasoning with a richer set of parameters, while lower-resource languages struggle with significant gaps.
Continuous reasoning in latent space crushes explicit reasoning for multilingual tasks, especially when training data is scarce.
Looping helps transformers think harder on math problems, while memory lets them remember more commonsense facts, and combining both beats simply scaling up layers.
