Chinese-Jev: Bringing System One Model to Chinese-Language Tasks

2026-09-29

AI Summary

This study addresses the insufficient decision accuracy of existing Jev models on Chinese tasks by constructing System One, an efficient decision-making model tailored for the Chinese language. Methodologically, it introduces a unified probabilistic objective transformation protocol and the CJ-Bench benchmark to enable cross-domain shared training. The model adopts a lightweight encoder-only architecture, combined with two-stage pre-training and fine-tuning alongside INT8 quantization to optimize inference efficiency. Experimental results demonstrate that this model surpasses closed-source counterparts by 1.24% in general-domain accuracy while achieving a 20-fold increase in inference speed. With mobile deployment latency reduced to approximately one second, the proposed approach effectively balances high precision with low latency.

Abstract

System One models such as Jev offer an efficient alternative to generative language models for tasks that require decisions rather than open-ended responses. However, existing Jev models exhibit limited Chinese-language decision accuracy, restricting their utility in both general and specialized settings. In this paper, we introduce Chinese-Jev, a System One model that addresses this gap through a unified data processing and training pipeline. Our data processing protocol converts heterogeneous Chinese-language annotations into probability targets over candidate options, enabling a shared training formulation across domains and question formats. To enable efficient inference, Chinese-Jev adopts a lightweight encoder-only backbone for text encoding and learns to score candidate answers through decision-oriented training. To address the misalignment between the pre-training distribution and downstream Chinese-language scenarios, we first train the model on a general-purpose corpus of 10 million examples, then fine-tune it separately for the medical, legal, and financial domains. To evaluate decision accuracy and calibration in both general and domain-specific Chinese-language settings, we introduce Chinese-Jev Bench (CJ-Bench). After first-stage pre-training, Chinese-Jev exceeds the accuracy of the closed-source Jev model by 1.24% on general-domain tasks while achieving a 20.3x speedup. Subsequent domain-specific fine-tuning yields a 4.0% accuracy improvement over Jev in medicine and achieves 92% of Jev's average accuracy across specialized domains, with a 17x speedup and an average latency of only 15 ms per example. We further demonstrate on-device deployment of an INT8-quantized model on mobile devices, achieving an inference latency of approximately 1.0 second per decision. The project is available at https://gulucaptain.github.io/Chinese-Jev/.

Problem

• System One models
• Chinese-language tasks
• decision accuracy
• Jev

Innovation

• System One model
• encoder-only architecture
• unified data processing pipeline
• domain-specific fine-tuning
• on-device deployment