Multi-turn interactions with large language models typically retain the assistant's own past responses in the conversation history. In this work, we revisit this design choice by asking whether large language models benefit from conditioning on their own prior responses. Using in-the-wild, multi-turn conversations, we compare standard (full-context) prompting with a user-turn-only prompting approach that omits all previous assistant responses, across three open reasoning models and one state-of-the-art model. To our surprise, we find that removing prior assistant responses does not affect response quality on a large fraction of turns. Omitting assistant-side history can reduce cumulative context lengths by up to 10x. To explain this result, we find that multi-turn conversations consist of a substantial proportion (36.4%) of self-contained prompts, and that many follow-up prompts provide sufficient instruction to be answered using only the current user turn and prior user turns. When analyzing cases where user-turn-only prompting substantially outperforms full context, we identify instances of context pollution, in which models over-condition on their previous responses, introducing errors, hallucinations, or stylistic artifacts that propagate across turns. Motivated by these findings, we design a context-filtering approach that selectively omits assistant-side context. Our findings suggest that selectively omitting assistant history can improve response quality while reducing memory consumption.

与大型语言模型的多轮交互通常会在对话历史记录中保留助手自己过去的响应。在这项工作中，我们通过询问大型语言模型是否受益于它们自己先前的响应来重新审视这种设计选择。使用野外多回合对话，我们将标准（全上下文）提示与用户仅回合提示方法进行比较，该方法忽略了所有先前的助理响应，跨越三种开放推理模型和一种最先进的模型。令我们惊讶的是，我们发现删除先前的助理响应并不会影响大部分转弯的响应质量。省略助手端历史记录可以将累积上下文长度减少多达 10 倍。为了解释这个结果，我们发现多轮对话包含很大比例（36.4％）的独立提示，并且许多后续提示提供了足够的指令，仅使用当前用户轮次和先前用户轮次即可回答。在分析仅用户轮次提示明显优于完整上下文的情况时，我们识别了上下文污染的实例，其中模型对其先前的响应进行了过度调节，引入了错误、幻觉或跨轮次传播的风格伪像。受这些发现的启发，我们设计了一种上下文过滤方法，有选择地忽略助手端上下文。我们的研究结果表明，有选择地省略助理历史记录可以提高响应质量，同时减少内存消耗。

GPU kernel optimization is fundamental to modern deep learning but remains a highly specialized task requiring deep hardware expertise. Despite strong performance in general programming, large language models (LLMs) remain uncompetitive with compiler-based systems such as torch.compile for CUDA kernel generation. Existing CUDA code generation approaches either rely on training-free refinement or fine-tune models within fixed multi-turn execution-feedback loops, but both paradigms fail to fundamentally improve the model's intrinsic CUDA optimization ability, resulting in limited performance gains. We present CUDA Agent, a large-scale agentic reinforcement learning system that develops CUDA kernel expertise through three components: a scalable data synthesis pipeline, a skill-augmented CUDA development environment with automated verification and profiling to provide reliable reward signals, and reinforcement learning algorithmic techniques enabling stable training. CUDA Agent achieves state-of-the-art results on KernelBench, delivering 100\%, 100\%, and 92\% faster rate over torch.compile on KernelBench Level-1, Level-2, and Level-3 splits, outperforming the strongest proprietary models such as Claude Opus 4.5 and Gemini 3 Pro by about 40\% on the hardest Level-3 setting.

GPU 内核优化是现代深度学习的基础，但仍然是一项高度专业化的任务，需要深厚的硬件专业知识。尽管在通用编程方面具有强大的性能，但大型语言模型 (LLM) 与基于编译器的系统（例如用于 CUDA 内核生成的 torch.compile）相比仍然没有竞争力。现有的 CUDA 代码生成方法要么依赖于免训练细化，要么依赖于固定多轮执行反馈循环内的微调模型，但这两种范式都无法从根本上提高模型内在的 CUDA 优化能力，导致性能提升有限。我们推出了 CUDA Agent，这是一个大规模代理强化学习系统，它通过三个组件开发 CUDA 内核专业知识：可扩展的数据合成管道、具有自动验证和分析功能的技能增强 CUDA 开发环境，以提供可靠的奖励信号，以及支持稳定训练的强化学习算法技术。 CUDA Agent 在 KernelBench 上实现了最先进的结果，在 KernelBench Level-1、Level-2 和 Level-3 分割上比 torch.compile 速度快 100%、100% 和 92%，在最难的 Level-3 设置上比 Claude Opus 4.5 和 Gemini 3 Pro 等最强的专有模型高出约 40%。