arXiv:2507.23170cs.LG2025-07
大模型服务难三全其美,预算、真实性和推理能力不可兼得。
BAR Conjecture: the Feasibility of Inference Budget-Constrained LLM Services with Authenticity and Reasoning
- 提出BAR定理,揭示三者间的不可兼得关系
- 证明在推理预算约束下无法同时保证真实与推理能力
- 为大模型应用设计提供理论指导,适合架构师参考
在设计大模型服务时,从业者关注三个核心属性:推理阶段预算、事实真实性与推理能力。然而我们的分析表明,没有模型能同时优化这三者。本文正式证明了这一权衡关系,并提出了名为BAR定理的系统性框架,用于指导大模型应用的设计。
原文摘要 · Abstract (English)
When designing LLM services, practitioners care about three key properties: inference-time budget, factual authenticity, and reasoning capacity. However, our analysis shows that no model can simultaneously optimize for all three. We formally prove this trade-off and propose a principled framework named The BAR Theorem for LLM-application design.
大模型推理优化系统设计
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。