Xuefei Ning

Uniform layer looping wastes latent compute on trivial tokens, but dynamically routing iteration depth beats standard transformers on AIME with a 53% steeper test-time scaling slope.

Isolated assessments may mask biases, but comparative evaluations can unleash hidden discrimination in LLMs, especially as model sizes grow.

Discrete diffusion models can be sped up by 14x by intelligently choosing which tokens to sample at each step, without sacrificing accuracy.