KAT-Coder Technical Report

  • 2025-10-21 16:27:47
  • Zizheng Zhan, Ken Deng, Xiaojiang Zhang, Jinghui Wang, Huaixi Tang, Zhiyi Lai, Haoyang Huang, Wen Xiang, Kun Wu, Wenhao Zhuang, Minglei Zhang, Shaojie Wang, Shangpeng Yan, Kepeng Lei, Zongxian Feng, Huiming Wang, Zheng Lin, Mengtong Li, Mengfei Xie, Yinghan Cui, Xuxing Chen, Chao Wang, Weihao Li, Wenqiang Zhu, Jiarong Zhang, Jingxuan Xu, Songwei Yu, Yifan Yao, Xinping Lei, Han Li, Junqi Xiong, Zuchen Gao, Dailin Li, Haimo Li, Jiaheng Liu, Yuqun Zhang, Junyi Peng, Haotian Zhang, Bin Chen
  • 0

Abstract

Recent advances in large language models (LLMs) have enabled progress inagentic coding, where models autonomously reason, plan, and act withininteractive software development workflows. However, bridging the gap betweenstatic text-based training and dynamic real-world agentic execution remains acore challenge. In this technical report, we present KAT-Coder, a large-scaleagentic code model trained through a multi-stage curriculum encompassingMid-Term Training, Supervised Fine-Tuning (SFT), Reinforcement Fine-Tuning(RFT), and Reinforcement-to-Deployment Adaptation. The Mid-Term stage enhancesreasoning, planning, and reflection capabilities through a corpus of realsoftware engineering data and synthetic agentic interactions. The SFT stageconstructs a million-sample dataset balancing twenty programming languages, tendevelopment contexts, and ten task archetypes. The RFT stage introduces a novelmulti-ground-truth reward formulation for stable and sample-efficient policyoptimization. Finally, the Reinforcement-to-Deployment phase adapts the modelto production-grade IDE environments using Error-Masked SFT and Tree-StructuredTrajectory Training. In summary, these stages enable KAT-Coder to achieverobust tool-use reliability, instruction alignment, and long-context reasoning,forming a deployable foundation for real-world intelligent coding agents. OurKAT series 32B model, KAT-Dev, has been open-sourced onhttps://huggingface.co/Kwaipilot/KAT-Dev.

 

Quick Read (beta)

loading the full paper ...