pith. sign in

DiLaDiff: Distilled Latent-Augmented Diffusion for Language Modeling

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it
abstract

Diffusion language models intrinsically fail to capture correlations between decoded tokens, which leads to a harsh trade-off between sampling quality and throughput. To solve this issue, we propose DiLaDiff, a variant of masked diffusion language models with three components: (1) a continuous latent space with semantic capabilities, learned by an auto-encoder fine-tuned from an existing masked diffusion language model; (2) a latent diffusion model learning the prior over the encoder distribution; (3) a consistency model distilling the learned prior into a few-step latent generative model. We show that, even without distillation, our latent-guided diffusion model outperforms the masked diffusion baseline while significantly accelerating inference. Consistency distillation further lowers the computational overhead of continuous diffusion, such that the latent is generated in negligible time compared to discrete decoding.

fields

cs.CL 1

years

2026 1

verdicts

UNVERDICTED 1

clear filters

representative citing papers

Continuous Language Diffusion as a Decoder-Interface Problem

cs.CL · 2026-06-07 · unverdicted · novelty 7.0

Continuous language diffusion works by entering high-margin decoder basins where frozen T5 embeddings recover 93-96% of native decisions and linear readouts reach 97.9% agreement, implying models should be evaluated as representation-decoder systems.

citing papers explorer

Showing 1 of 1 citing paper after filters.

  • Continuous Language Diffusion as a Decoder-Interface Problem cs.CL · 2026-06-07 · unverdicted · none · ref 35 · internal anchor

    Continuous language diffusion works by entering high-margin decoder basins where frozen T5 embeddings recover 93-96% of native decisions and linear readouts reach 97.9% agreement, implying models should be evaluated as representation-decoder systems.