AI科学家竞赛应聚焦人机协作而非单智能体
The race to build “AI Scientists” may be optimizing for the wrong unit: the agen…
The race to build “AI Scientists” may be optimizing for the wrong unit: the agent alone.
构建“AI科学家”的竞赛可能正在优化错误的单元：仅关注智能体本身。
This position paper argues that scientific agents should be studied as human-agent systems, where the thing you evaluate is the scientist + agent pair.
本立场论文主张，科学智能体应作为人机系统来研究，其中评估的对象是科学家与智能体的组合。
Most current systems still treat the human as a supervisor: set the goal, review a phase, approve the final artifact. Far fewer are built for continuous, fine-grained collaboration during the work itself.
当前大多数系统仍将人类视为监督者：设定目标、审查阶段、批准最终成果。很少有系统设计用于在工作过程中进行持续、细粒度的协作。
The problem is that agents do not naturally know when they need human input. In 10 science tasks, GPT-5-mini almost never asked for help.
问题在于，智能体并不自然知道何时需要人类输入。在10个科学任务中，GPT-5-mini几乎从不主动请求帮助。
But that input matters: in the case studies, experts caught errors the agents missed, while the agents sped up execution. The gain came from collaboration, not autonomy.
但该输入至关重要：在案例研究中，专家发现了智能体遗漏的错误，而智能体则加速了执行过程。收益来自协作，而非自主性。
The paper’s proposed benchmark is therefore different: does the human-agent team produce better science than either member alone, without collaboration cost overwhelming the gain?
因此，论文提出的基准测试有所不同：人机团队是否比任何一方单独工作产生更好的科学成果，且协作成本不压倒收益？
– arxiv. org/abs/2608.14667
– arxiv.org/abs/2608.14667
Title: "Position: AI Agents in Scientific Teams Should Be Studied as Human-Agent Systems"
标题：“立场：科学团队中的AI智能体应作为人机系统来研究”
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力