NCP-ArchPreview Technical Report: Moving towards Latent Space Language Models through Next Concept Prediction

Abstract

We introduce NCP-ArchPreview, a latent-space language model that pushes autoregressive pretraining beyond standard next-token prediction (NTP). Alongside NTP, the model learns through Next Concept Prediction (NCP) to predict discrete concepts that span multiple tokens, introducing an explicit and more challenging concept-level objective while preserving standard token-level autoregressive generation. NCP-ArchPreview builds a latent space by constructing a product-quantized concept vocabulary directly from its hidden states, and subsequently learns to predict future concepts via a dedicated Concept Module. These predicted concepts are then fed back to the token level to guide subsequent generation, with NTP and NCP trained jointly end-to-end. We scale this architecture to 8.9B parameters and train it on 5.73T tokens from the Dolma-3 dataset, marking the largest demonstration of a latent-space language model to date. Remarkably, by consuming only 51.3% of the total training tokens, NCP-ArchPreview achieves the final pretraining loss of OLMo-3-7B. Following full pretraining, it outperforms OLMo-3-7B by 2.45 points on the downstream macro-average, including a notable 5.99-point gain on GSM8K. Controlled experiments isolate a clear progression of performance gains stemming from both the latent architecture and the NCP objective. Furthermore, utilizing only 85% of the standard computation, NCP-ArchPreview approaches the training loss of a strictly parameter-aligned 8.9B baseline. The learned latent space remains highly valuable after the pretraining stage: updating just the 17M-parameter VQ module yields a novel, lightweight interface for domain adaptation, while a simple injection of concept representations into a DFlash2 drafter improves the mean accepted length by 4.17% with negligible overhead.

Authors 28

NCP Team, Jiaqi Cao, Chiyu Chen, Shuang Cheng, Xu Cheng, Beiya Dai, Yufan Feng, Kewen Ge, Ruijun Ge, Jiayi Huang, Yang Jiao, Dahua Lin, Zhouhan Lin, Yifan Liu, Yuliang Liu, Biqing Qi, Mowen Ruan, Junzhe Shen, Yunchong Song, Hao Sun, Zhongbo Tian, Yixuan Wang, Rubin Wei, Jiaxin Xiong, Kangyu Yang, Qian Yao, Qi Zhang, Bowen Zhou

Specification

Official page

arxiv.org/abs/2609.10715

Source:arXiv (Atom API + RSS)·T1·observed 5 h ago·high

Arxiv announce type

new

Source:arXiv (Atom API + RSS)·T1·observed 5 h ago·high

arXiv id

2609.10715

Source:arXiv (Atom API + RSS)·T1·observed 5 h ago·high

Categories

cs.CL

Source:arXiv (Atom API + RSS)·T1·observed 5 h ago·high

Hf paper url

https://huggingface.co/papers/2609.10715

Source:Hugging Face Hub (public pages, model cards, papers)·T2·observed 5 h ago·medium

Hf comments

2

Source:Hugging Face Hub (public pages, model cards, papers)·T2·observed 49 min ago·medium

Upvotes

177

Source:Hugging Face Hub (public pages, model cards, papers)·T2·observed 49 min ago·medium

PDF

arxiv.org/pdf/2609.10715

Source:arXiv (Atom API + RSS)·T1·observed 5 h ago·high

Primary category

cs.CL

Source:arXiv (Atom API + RSS)·T1·observed 5 h ago·high

Published

11 Sept 2026

Source:arXiv (Atom API + RSS)·T1·observed 5 h ago·high

Each value shows its source, tier and observation time. Conflicting claims are kept side by side and flagged — never averaged. How AI Atlas records facts →

Provenance