AURORA-LM: Autoencoding Unified Representation for Continuous-Latent Diffusion Language Modeling
2026-08-03 • Computation and Language
Computation and Language
AI summaryⓘ
The authors present AURORA-LM, a new language model that generates text using continuous latent spaces rather than traditional discrete tokens. They create a special way to represent text that is easy to decode and design a diffusion model that directly learns the distribution of these continuous text latents. Their approach, including block-wise generation and noise calibration, improves text generation quality without sacrificing detail. Testing shows AURORA-LM performs better than other continuous and diffusion-based language models on tasks like free text generation and summarization.
continuous latent spacediffusion modeltext generationautoencoderlatent representationflow matchingQuery-based Encoder-DecoderBlock-causal Transformernoise calibrationsummarization
Authors
Jiajun Liang, Yucheng Liao, Yukang Cao, Jiazhe Wei, Ken Li, Wende Tan, Jiankun Zhang, ZY Cui, Jingkang Yang, Liucheng Guo, Shiqi Yang, B. Yang, Caifeng Shan, Ziwei Liu, Chenyang Si
Abstract
Language remains an outlier in generative modeling: while images, video, and audio are increasingly modeled in continuous latent spaces, text generation still relies predominantly on discrete tokens. Existing continuous language models either inherit embedding spaces not designed for joint generation and decoding, or compress autoencoded latents to ease diffusion, sacrificing token-level fidelity. Instead of simplifying the representation to suit the generative model, we preserve a high-capacity, decodable text latent and design the diffusion model to learn its distribution directly. We introduce AURORA-LM, a continuous-latent diffusion language model that separates the construction of a decodable text representation from the modeling of its distribution. A Query-based Encoder-Decoder organizes text into a high-capacity, prefix-aligned latent sequence, and a Block-causal Diffusion Transformer learns its distribution through flow matching, generating blocks left to right while denoising positions within each block in parallel. Because such a latent is harder for diffusion to model, AURORA-LM restricts only the noisy-input pathway while retaining the full clean-latent prediction target, accommodating full-width latents without reducing decoder-facing capacity. We further calibrate the noise-level distribution to the latent width, and introduce self-trajectory consistency to bridge independently sampled training noise and iterative denoising at inference. AURORA-LM achieves the strongest performance among evaluated continuous and diffusion-based language models on OpenWebText free generation and XSum summarization. Scaling to 1B parameters with about 1500 EFLOPs of total compute yields further gains, surpassing a larger publicly released latent-diffusion language model under a matched evaluation protocol. All experiments are conducted on Ascend NPUs.