AReaL-Ascend v1.0.5 enables stable 64K-context Code Agent RL for Qwen3.6-27B on Ascend NPUs, using THD Sequence Packing to eliminate padding waste and MTP speculative decoding to accelerate rollout, achieving reward improvement from 0.56 to 0.94 over 120 steps with train-rollout consistency near 1.0.
This article traces Claude Code agent interactions to reveal how natural language queries are converted into search commands via agentic tool-driven loops, highlighting the impact of tool configuration and LLM reasoning quality on code location accuracy.
CoDA‑Bench, a new benchmark from RUC, places code agents in a sandbox containing over a thousand heterogeneous data files and requires them to locate the correct dataset, write analysis code, and produce answers, revealing that current agents achieve only about 61 % accuracy overall and struggle mainly with data discovery rather than code generation.
The author demonstrates using Claude Code agent to automatically download Android 16 and 17 source code via repo init --reference for space efficiency, then generate an HTML diff report comparing simpleperf repository changes with side-by-side visualization and feature summaries, replacing manual Beyond Compare workflows.
In a live discussion hosted by Prof. Yang Jian with experts Zhang Xi and Cui Chen, the panel explores the essential abilities of mature GUI agents, the role of multimodal models in visual understanding, the transfer of code‑agent techniques to GUI tasks, edge‑device performance trade‑offs, complex planning, tool ecosystems, deployment challenges, and future breakthrough scenarios.
A round‑table of AI experts and industry leaders examines the current capabilities, limitations, and future trajectories of code agents, covering topics from capability boundaries and autonomous evolution to large‑scale codebase challenges, multi‑agent collaboration, hallucination mitigation, and security safeguards.
This article describes how the Gaode terminal team tackled large‑scale repository upgrades by building a code‑RAG and code‑Agent tool, addressing recall and stability issues, then fine‑tuning a small LLM (Qwen3‑4B) with LoRA and custom datasets to achieve reliable, low‑cost, on‑device code‑query performance.
