arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2603.16733cs.AIcs.CLcs.SE
收藏

IQuest-Coder-V1 技术报告

IQuest-Coder-V1 Technical Report

IQuest Coder Team(IQuest Coder团队)
Jian Yang, Wei Zhang, Shawn Guo, Zhengmao Ye, Lin Jing, Shark Liu, Yizhi Li, Jiajun Wu, Cening Liu, X. Ma, Yuyang Song, Siwei Wu, Yuwen Li, L. Liao, T. Zheng, Z… 展开作者

Jian Yang, Wei Zhang, Shawn Guo, Zhengmao Ye, Lin Jing, Shark Liu, Yizhi Li, Jiajun Wu, Cening Liu, X. Ma, Yuyang Song, Siwei Wu, Yuwen Li, L. Liao, T. Zheng, Ziling Huang, Zelong Huang, Che Liu, Yan Xing, Renyuan Li, Qingsong Cai, Hanxu Yan, Siyue Wang, Shikai Li, Jason Klein Liu, An Huang, Yongsheng Kang, Jinxing Zhang, Chuan Hao, Haowen Wang, Weicheng Gu, Ran Tao, Mingjie Tang, Peihao Wu, Jianzhou Wang, Xianglong Liu, Weifeng Lv, Bryan Dai

AI总结:

本文介绍IQuest-Coder-V1系列模型,通过代码流多阶段训练范式提升软件逻辑动态演化能力,结合预训练、中训练和后训练阶段,实现代码智能在代理软件工程、编程竞赛和复杂工具使用中的先进性能。

摘要:

本报告介绍了IQuest-Coder-V1系列(7B/14B/40B/40B-Loop),一种新的代码大语言模型(LLMs)家族。超越静态代码表示,我们提出了代码流多阶段训练范式,通过管道的不同阶段捕捉软件逻辑的动态演变。我们的模型通过进化管道开发,从包含代码事实、仓库和完成数据的初始预训练开始。随后,我们实现了一个专门的中训练阶段,整合推理和代理轨迹,在32k上下文和128k上下文规模上构建深厚的逻辑基础。然后,通过后训练专门的编码能力,将其分为两条专用路径:思考路径(利用推理驱动的强化学习)和指导路径(优化通用帮助)。IQuest-Coder-V1在代码智能的关键维度上实现了最先进的性能:代理软件工程、编程竞赛和复杂工具使用。为了解决部署限制,IQuest-Coder-V1-Loop变种引入了递归机制,以优化模型容量和部署足迹之间的权衡,提供了一条在效能-效率权衡上的架构增强路径。我们相信IQuest-Coder-V1系列的发布,包括从预训练基础到最终思考和指导模型的完整白盒链,将推动自主代码智能和现实世界代理系统的研究。

英文摘要:

In this report, we introduce the IQuest-Coder-V1 series-(7B/14B/40B/40B-Loop), a new family of code large language models (LLMs). Moving beyond static code representations, we propose the code-flow multi-stage training paradigm, which captures the dynamic evolution of software logic through different phases of the pipeline. Our models are developed through the evolutionary pipeline, starting with the initial pre-training consisting of code facts, repository, and completion data. Following that, we implement a specialized mid-training stage that integrates reasoning and agentic trajectories in 32k-context and repository-scale in 128k-context to forge deep logical foundations. The models are then finalized with post-training of specialized coding capabilities, which is bifurcated into two specialized paths: the thinking path (utilizing reasoning-driven RL) and the instruct path (optimized for general assistance). IQuest-Coder-V1 achieves state-of-the-art performance among competitive models across critical dimensions of code intelligence: agentic software engineering, competitive programming, and complex tool use. To address deployment constraints, the IQuest-Coder-V1-Loop variant introduces a recurrent mechanism designed to optimize the trade-off between model capacity and deployment footprint, offering an architecturally enhanced path for efficacy-efficiency trade-off. We believe the release of the IQuest-Coder-V1 series, including the complete white-box chain of checkpoints from pre-training bases to the final thinking and instruction models, will advance research in autonomous code intelligence and real-world agentic systems.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑