揭秘架构如何成为编码智能体的能力均衡器，为提升代码生成质量提供新视角。
arXiv:2608.21747v1 Announce Type: cross Abstract: LLM-based coding agents generate complete software systems from high-level descriptions, yet little …
长时任务中让智能体“记性好”的新方案：PRO-LONG 用程序化记忆显著提升编码表现，费用还省数倍。
Article URL: https://github.com/alexisfox7/PRO-LONG Comments URL: https://news.ycombinator.com/item?id=49329010 Points: 1 # Comments: 0
匹配分数会掩盖命令路径失败？QuoteBench用最终状态验证揭示LLM编码代理的真实边界。
arXiv:2608.13547v1 Announce Type: new Abstract: LLM coding agents issue Bash commands through interfaces that may serialize, wrap, and reparse model o…
基于llamafile/GGUF的终端编码智能体，本地运行高效又灵活
Article URL: https://github.com/kdeps/kdeps Comments URL: https://news.ycombinator.com/item?id=49029939 Points: 2 # Comments: 0
4B小模型能否硬刚前沿大模型？专攻智能体执行任务，效率与能力兼得的新答案来了。
arXiv:2605.03195v2 Announce Type: replace Abstract: Modern coding agents increasingly delegate specialized subtasks to subagents, which are smaller, f…
用BitTorrent思路把闲置编码智能体连成P2P网格，零API跑本地小模型，适合分布式任务派发与算力共享。
Article URL: https://github.com/raghavan/agent-torrent Comments URL: https://news.ycombinator.com/item?id=48793521 Points: 3 # Comments: 0
首个编码智能体负载画像公开数据集，揭示LLM服务系统的长上下文与爆发性流量特征，优化推理部署必读。
arXiv:2606.30560v1 Announce Type: new Abstract: Coding agents are rapidly becoming a major application of agentic LLMs, but serving them efficiently r…
贝叶斯控制理论为代码生成代理注入概率推理，提升行为可靠性与稳定性
arXiv:2606.24453v1 Announce Type: new Abstract: Modern coding agents pair LLM generators with various tools, including cheap diagnostics and expensive…
编码代理忽略代码结构？这项研究提出结构索引方案，让AI更懂代码库。
arXiv:2606.22417v1 Announce Type: new Abstract: Coding agents now interleave LLMs with retrieval over the working repository, and retrieval implementa…
让编码智能体读懂仓库的“潜规则”？Probe-and-Refine调优法挖掘AGENTS.md指导价值，显著提升代码修复准确率。
arXiv:2606.20512v1 Announce Type: cross Abstract: LLM-based coding agents need higher-level operational knowledge about a repository (which files hous…
别只换模型，团队用 LLM 的差距在“驾驶”技巧。看工程师如何把 Claude 变成高效编码搭档。
Article URL: https://home.robusta.dev/blog/how-to-drive-an-llm Comments URL: https://news.ycombinator.com/item?id=48596318 Points: 3 # Comments: 0
提出可执行世界模型方法，在编码代理时代攻克ARC-AGI-3挑战，为通用人工智能探索新路径。
arXiv:2605.05138v2 Announce Type: replace Abstract: We evaluate an initial coding-agent system for ARC-AGI-3 in which the agent maintains an executabl…
首个针对AI编码智能体的安全基准测试，助你评估模型在代码生成中的安全防护能力。
Article URL: https://gist.github.com/allenwu-blip/fa2bd0218b93a1d7aef765817e3c6608 Comments URL: https://news.ycombinator.com/item?id=48274727 Points:…
深入评估编码智能体在数据驱动科学发现中的记忆压缩策略，为AI辅助科研提供新思路。
arXiv:2605.18854v1 Announce Type: new Abstract: Coding agents accumulate extensive context during long-running tasks, yet fixed context windows force …
全新基准测试聚焦代码智能体在连续软件包版本升级中的能力，比单次任务更贴近真实维护场景。
arXiv:2605.14415v1 Announce Type: cross Abstract: Coding agents powered by large language models are increasingly expected to perform realistic softwa…
一种通过代码生成构建物理世界模型的智能体框架，利用规划、编码、视觉审查和物理分析代理迭代修正，确保模拟的物理真实性和指令符合度，首次从底层解决了视频世界模型“有形无实”的缺陷。
arXiv:2605.14398v1 Announce Type: new Abstract: World models have emerged as a powerful paradigm for building interactive simulation environments, wit…