PulseAugur
中
实时 21:34:11
English(EN) Same model, same task, 2.6x the bill: where agent harness costs actually go

AI 智能体(agent)的“挽具”(harness)设计显著影响成本,而非准确性

来自加州大学伯克利分校、西北大学和伊利诺伊大学厄巴纳-香槟分校的一项新研究论文揭示,“挽具”(harness)——即围绕大型语言模型(LLM)的系统提示、工具和循环——在成本上对 AI 智能体(agent)有显著影响,但对准确性没有影响。研究发现,改变“挽具”可能使相同模型和任务的成本增加高达 2.6 倍,这主要是由于在无状态 LLM 调用中重复重新读取上下文。研究人员分析了 Claude Code、Codex 和 GPT-6 Luna 等模型的运行轨迹,观察到尽管不同“挽具”的准确性相似,但更复杂的“挽具”会产生显著更高的费用,这通常是由于冗长的系统提示和工具模式。 AI

影响 强调优化 LLM 智能体(agent)的“挽具”(harness)可以在不牺牲性能的情况下显著降低运营成本。

排序理由 关于 LLM 智能体(agent)“挽具”(harness)成本影响的研究论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 dev.to — LLM tag 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →

AI 智能体(agent)的“挽具”(harness)设计显著影响成本,而非准确性

本文如何被排名

Signal score
30 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
关于 LLM 智能体(agent)“挽具”(harness)成本影响的研究论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
infra, product
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

  1. dev.to — LLM tag TIER_1 English(EN) · Ahmad Dogar ·

    同一模型,同一任务,账单却高出2.6倍:智能体部署成本的真正去向

    <p>A coding agent session that runs for an hour does not cost twelve times what a five minute one does. It costs a lot more than that, and the reason is not the model getting chattier. It is the shape of the loop: every turn sends the whole conversation back to the model, and you…