Searcharxiv⌕ Search

arXiv subjects

Jiarong Zhang

Publications and source records attributed to Jiarong Zhang.

4 recordsLinked to original sources
arXiv2025

KAT-Coder Technical Report

Recent advances in large language models (LLMs) have enabled progress in agentic coding, where models autonomously reason, plan, and act within interactive software development workflows. However, bridging the gap between static text-based training and dynamic real-world agentic execution remains a core challenge. In this technical report, we present KAT-Coder, a large-scale agentic code model trained through a multi-stage curriculum encompassing Mid-Term Training, Supervised Fine-Tuning (SFT), Reinforcement Fine-Tuning (RFT), and Reinforcement-to-Deployment Adaptation. The Mid-Term stage enhances reasoning, planning, and reflection capabilities through a corpus of real software engineering data and synthetic agentic interactions. The SFT stage constructs a million-sample dataset balancing twenty programming languages, ten development contexts, and ten task archetypes. The RFT stage introduces a novel multi-ground-truth reward formulation for stable and sample-efficient policy optimization. Finally, the Reinforcement-to-Deployment phase adapts the model to production-grade IDE environments using Error-Masked SFT and Tree-Structured Trajectory Training. In summary, these stages enable KAT-Coder to achieve robust tool-use reliability, instruction alignment, and long-context reasoning, forming a deployable foundation for real-world intelligent coding agents. Our KAT series 32B model, KAT-Dev, has been open-sourced on https://huggingface.co/Kwaipilot/KAT-Dev.

cs.CL↗
arXiv2025

SeamlessFlow: A Trainer Agent Isolation RL Framework Achieving Bubble-Free Pipelines via Tag Scheduling

We introduce SeamlessFlow, a server based reinforcement learning (RL) framework that addresses two core challenges in industrial scale RL: (1) decoupling RL training from the complex execution flow of agents; (2) maximizing GPU utilization with minimal idle time while preserving the stability and scalability required for large-scale deployments. First, SeamlessFlow introduces a data plane that decouples the RL trainer from diverse, complex agent implementations while sustaining high throughput. A central trajectory manager maintains complete interaction histories and supports partial rollout, allowing rollout to pause for weight updates and resume seamlessly, keeping agents unaware of service interruptions. Second, we propose a tag driven scheduling paradigm that abstracts hardware into capability tagged resources, unifying colocated and disaggregated architectures. Based on this, SeamlessFlow introduces a spatiotemporal multiplexing pipeline that dynamically reassigns idle training nodes to rollout in a train rollout separated setup, eliminating pipeline bubbles and fully exploiting heterogeneous cluster resources. By combining these innovations, SeamlessFlow delivers both stability and high performance, making it well suited for multi agent, long horizon, and other complex RL tasks.

cs.LG↗
arXiv2025

Constructions and List Decoding of Sum-Rank Metric Codes Based on Orthogonal Spaces over Finite Fields

Sum-rank metric codes, as a generalization of Hamming codes and rank metric codes, have important applications in fields such as multi-shot linear network coding, space-time coding and distributed storage systems. The purpose of this study is to construct sum-rank metric codes based on orthogonal spaces over finite fields, and calculate the list sizes outputted by different decoding algorithms. The following achievements have been obtained. In this study, we construct a cyclic orthogonal group of order $q^n-1$ and an Abelian non-cyclic orthogonal group of order $(q^n-1)^2$ based on the companion matrices of primitive polynomials over finite fields. By selecting different subspace generating matrices, maximum rank distance (MRD) codes with parameters $(n \times {2n}, q^{2n}, n)_q$ and $(n \times {4n}, q^{4n}, n)_q$ are constructed respectively. Two methods for constructing sum-rank metric codes are proposed for the constructed MRD codes, and the list sizes outputted under the list decoding algorithm are calculated. Subsequently, the $[{\bf{n}},k,d]_{{q^n}/q}$-system is used to relate sum-rank metric codes to subspace designs. The list size of sum-rank metric codes under the list decoding algorithm is calculated based on subspace designs. This calculation method improves the decoding success rate compared with traditional methods.

cs.IT↗