自动搜索最优推理技术组合，让大模型在有限算力下表现更优。
Archon: An Architecture Search Framework for Inference-Time Techniques
- 模块化框架自动探索多种推理技术的组合方式
- 在多任务上平均提升15.1%性能，优于主流模型
- 适合需要高效利用算力提升模型表现的研究者
推理时技术（如重复采样或迭代修正）正成为增强大语言模型（LLMs）测试阶段性能的有效手段。然而，由于对各类技术在不同模型与任务中的效用、彼此间交互关系以及组合配置的庞大搜索空间缺乏理解，相关系统设计仍不成熟。为此，我们提出 Archon——一个模块化且自动化的框架，用于优化推理时技术与 LLM 的选择和组合。在给定计算预算和可用模型集的前提下，Archon 能探索大规模设计空间，生成针对目标基准的优化配置。其可设计定制化或通用架构，在准确率与最大令牌预算的权衡上超越现有先进基线。在指令遵循、推理和编程任务中，相比 OpenAI o1、GPT-4o 及 Claude 3.5 Sonnet 等前沿模型，平均性能提升 15.1%。
原文摘要 · Abstract (English)
Inference-time techniques, such as repeated sampling or iterative revisions, are emerging as powerful ways to enhance large-language models (LLMs) at test time. However, best practices for developing systems that combine these techniques remain underdeveloped due to our limited understanding of the utility of each technique across models and tasks, the interactions between them, and the massive search space for combining them. To address these challenges, we introduce Archon, a modular and automated framework for optimizing the process of selecting and combining inference-time techniques and LLMs. Given a compute budget and a set of available LLMs, Archon explores a large design space to discover optimized configurations tailored to target benchmarks. It can design custom or general-purpose architectures that advance the Pareto frontier of accuracy vs. maximum token budget compared to top-performing baselines. Across instruction-following, reasoning, and coding tasks, we show that Archon can leverage additional inference compute budget to design systems that outperform frontier models such as OpenAI's o1, GPT-4o, and Claude 3.5 Sonnet by an average of 15.1%.
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。