Cover Image

Saved in:
Bibliographic Details
Main Authors: AGI Team, Cai, Yuxuan, Chen, Lu, Chen, Qiaoling, Ding, Yuyang, Fan, Liwen, Fu, Wenjie, Gao, Yufei, Guo, Honglin, Guo, Pinxue, Han, Zhenhua, He, Zhengfu, Hu, Hanglei, Hu, Kai, Hua, Shengjia, Huai, Tianyu, Huang, Baodai, Ji, Li, Jiang, Zhen, Lei, Zhikai, Li, Bufan, Lin, Jiahang, Lin, Lizhi, Liu, Jinxiu, Liu, Shichun, Liu, Ziming, Ni, Yuchen, Qian, Pengfang, Shen, Yujiong, Shi, Qingyun, Shu, Wentao, Sun, Peng, Suo, Yiran, Tang, Tian, Tian, Boyu, Wang, Guoteng, Wang, Junzhe, Wang, Peixin, Xi, Zhiheng, Yan, Hang, Yang, Jie, Yang, Zhixiong, Yao, Tianchu, Ye, Guangze, Yu, Qianxi, Zhang, Shuo, Zhang, Xinyue, Zhang, Yiqi, Zhao, Jiarong, Zheng, Miao, Zheng, Rui, Zhou, Enyu, Zhou, Jiazheng, Zhou, Maosen, Zhou, Yuhao, Gui, Tao, Zheng, Yining, Chen, Xinchi, Zhou, Jie, Feng, Siyuan, Chen, Qin, He, Liang, Zhang, Qi, Huang, Xuanjing, Qiu, Xipeng
Format: Preprint
Published: 2025
Subjects:
Computation and Language
Online Access:https://arxiv.org/abs/2512.04987
Tags: Add Tag
No Tags, Be the first to tag this record!
Table of Contents:
  • The evolution of Large Language Models (LLMs) from passive responders to autonomous agents necessitates a fundamental shift in learning paradigms -- from static imitation to incentive-driven decision making. However, this transition is significantly impeded by the lack of scalable infrastructure capable of constructing high-quality interaction signals for effective policy learning. To address this, we introduce a comprehensive method designed to systematically scale the diversity and complexity of interactive environments. Our method realizes this scaling by addressing three orthogonal dimensions: (1) Complexity: NexAU, a flexible agent framework that supports building complex agent hierarchies via simple configurations; (2) Diversity: NexA4A automatically generates diverse agent hierarchies from natural language to cover infinite domains; and (3) Fidelity: NexGAP bridges the simulation-reality gap by integrating dynamic real-world environment for grounded trajectories synthesis. We train Nex-N1 upon the diverse and complex interactive environments established by our infrastructure. Empirical results on benchmarks such as SWE-bench and tau2 demonstrate that Nex-N1 consistently outperforms SOTA open-source models and achieves competitive performance against frontier proprietary models on complex agentic tasks. We open-source the Nex ecosystem and model weights to facilitate further research.

Similar Items