微软论文：AI谈判代理因过度礼貌常失败，SocialRL训练小模型媲美GPT
New Microsoft paper shows an AI agent that negotiates for you will usually lose,…
Agent落地谈判场景的核心痛点是利益保护，这篇论文给出了具体的SocialRL训练方案和小模型验证数据，对构建自主决策Agent有直接参考价值。
New Microsoft paper shows an AI agent that negotiates for you will usually lose, because it was trained to be agreeable.
微软最新论文表明，代表你进行谈判的 AI 智能体通常会失败，因为它被训练得过于顺从。
Politeness, transparency and eagerness to close are great in a chat assistant and terrible in a delegate. The paper found frontier models leaking their user's budget and folding the moment a seller pushed back.
礼貌、透明和急于成交在聊天助手身上是优点，但在代理人身上则是致命缺陷。论文发现，前沿模型会泄露用户的预算，并在卖家稍加施压时立即妥协。
Their fix is SocialRL: instead of prompting the model to negotiate better, train it on the outcome of the deal across six bargaining and scheduling games.
他们的解决方案是 SocialRL：与其通过提示让模型更好地谈判，不如在六种议价和调度游戏的交易结果上对其进行训练。
It works, and it doesn't take a big model. A 4B model started anchoring low, holding its position and walking away from bad deals, and landed at 0.627 average across all six games, matching GPT-4.1 at 0.625.
该方法有效，且不需要大模型。一个 4B 参数量的模型开始压低初始报价、坚守立场并拒绝糟糕的交易，在全部六款游戏中的平均得分达到 0.627，与 GPT-4.1 的 0.625 持平。
The catch is that prompting alone made things worse, so this is a training fix, not a prompt fix.
关键在于，仅靠提示反而会让情况变得更糟，因此这是一种训练层面的修复，而非提示层面的修复。
So if you're building an agent that acts on someone's behalf, stop scoring it on whether the deal closed and start scoring it on what it gave away.
因此，如果你正在构建代表他人行事的智能体，请停止以交易是否达成作为评分标准，转而以其让步程度作为评分依据。
– arxiv. org/abs/2608.13787
– arxiv.org/abs/2608.13787
Title: "From Passive Delegates to Strategic Negotiators: Reinforcing Social Reasoning in Small Language Models with SocialRL"
标题：《从被动代理人到战略谈判者：利用 SocialRL 强化小型语言模型中的社会推理能力》
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