让代码生成模型像用钱一样控制思考深度，又快又准。
Reasoning as a Resource: Optimizing Fast and Slow Thinking in Code Generation Models
- 把思考过程当资源调度，可快可慢
- 能提升准确率、降低延迟和成本
- 适合需要高效安全部署的开发者
本文提出一种代码生成模型设计的根本性转变：将推理深度视为可调控的资源。不再将快速回答（‘快思考’）与逐步推理解答（‘慢思考’）视为提示工程的附带结果，而是主张在模型全生命周期中显式管理二者权衡。通过优化推理预算——从合成数据生成、基准测试到实际部署——可实现准确率、延迟与成本间的更优平衡。该论文指出，对推理过程的自适应控制能增强监督信号，推动多维度新基准的建立，并指导成本敏感、安全优先的部署策略。将快思考与慢思考视作可调度的互补模式，我们展望未来编码代理可在必要时深入思考，也可在合适时迅速行动。
原文摘要 · Abstract (English)
This position paper proposes a fundamental shift in designing code generation models: treating reasoning depth as a controllable resource. Rather than being an incidental byproduct of prompting, we argue that the trade-off between rapid, direct answers ("fast thinking") and elaborate, chain-of-thought deliberation ("slow thinking") must be explicitly managed. We contend that optimizing reasoning budgets across the entire model lifecycle - from synthetic data creation and benchmarking to real-world deploymen - can unlock superior trade-offs among accuracy, latency, and cost. This paper outlines how adaptive control over reasoning can enrich supervision signals, motivate new multi-dimensional benchmarks, and inform cost-aware, security-conscious deployment policies. By viewing fast and slow thinking as complementary modes to be scheduled, we envision coding agents that think deep when necessary and act fast when possible.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。