Search papers, labs, and topics across Lattice.
3
0
6
Uniform-state diffusion models can outperform masked diffusion models in blockwise generation, challenging prior assumptions about their relative effectiveness.
Exponentially decaying memory can significantly boost the accuracy of existing sparse attention methods, suggesting a simple way to improve long-context LLM efficiency.
Uniform-state diffusion models can now achieve state-of-the-art generative performance in language modeling, thanks to a new Predictor-Corrector sampler that breaks the quality plateau of ancestral sampling.
