ArcticSwarm：通过搜索隔离解决多智能体过早共识问题
Letting research agents talk too early can make them follow the same wrong idea,…
针对多智能体协作中常见的“过早共识”痛点给出了具体的架构解法与量化对比，对构建高可靠性 Research Agent 有直接参考价值。
Letting research agents talk too early can make them follow the same wrong idea, so isolate some searches before review.
让研究代理过早交流可能导致它们遵循相同的错误思路，因此在审查前隔离部分搜索。
Multi-agent research works better when agents search independently before comparing notes, so delay collaboration until there is evidence to review.
多代理研究在代理独立搜索后再比对结果时效果更好，因此应推迟协作，直到有证据可供审查。
The problem is once 1 agent finds a plausible answer, other agents can start searching around that same idea instead of testing different possibilities.
问题在于，一旦一个代理找到了看似合理的答案，其他代理可能会围绕该相同思路进行搜索，而不是测试不同的可能性。
The paper calls this premature consensus.
论文将这种现象称为“过早共识”。
ArcticSwarm fixes it by blocking selected agents from reading their peers while they search, then bringing the findings together for review.
ArcticSwarm 通过阻止选定的代理在搜索期间阅读同伴的信息来修复此问题，然后将发现汇总以供审查。
On BrowseComp-Plus with Qwen 3.5-27B, it reached 82.6% accuracy.
在 BrowseComp-Plus 上使用 Qwen 3.5-27B，其准确率达到 82.6%。
Remove that search isolation and accuracy fell to 78.8%.
移除搜索隔离后，准确率降至 78.8%。
Remove the review system too, and it fell to 74.5%.
同时移除审查系统后，准确率进一步降至 74.5%。
Even 40 independent single-agent runs with majority voting reached only 63.5%.
即使进行 40 次独立的单代理运行并采用多数投票，准确率也仅为 63.5%。
So the lesson is not "add more agents" or "make them communicate more."
因此，教训不是“增加更多代理”或“让它们更多地交流”。
For difficult research tasks without a reliable verifier, give agents room to explore different paths first, then challenge and verify the leading answer before the swarm agrees.
对于缺乏可靠验证器的困难研究任务，先给予代理探索不同路径的空间，然后在蜂群达成一致之前挑战和验证领先答案。
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