做多代理系统或谈判 AI 的开发者，这篇论文提供了一个实用的对手建模框架——用有限交互预测对方决策，比直接提示更准。建议点开看看他们的 LLM-as-Observer 方法，可能帮你省掉大量试错成本。
该研究探讨了 AI 代理如何在有限交互中预测陌生对手的下一步决策，例如买家代理面对未知卖家或采购助理与供应商谈判。研究通过受控谈判游戏将问题建模为目标自适应文本表格预测，每个决策点结合结构化游戏状态、报价历史和对话。模型基于表格基础模型，并引入 LLM-as-Observer 作为额外表示层：冻结的小型 LLM 读取决策时刻状态和对话，其隐藏状态作为决策导向特征，而非直接预测。在 13 个前沿 LLM 代理上训练并测试 91 个保留代理，该模型在响应预测 AUC 上提升约 4 点，报价预测误差降低 14%，证明隐藏的 LLM 表示能暴露直接提示无法获取的决策信号。
Predicting Decisions of AI Agents from Limited Interaction through Text-Tabular Modeling
AI agents negotiate and transact in natural language with unfamiliar counterparts: a buyer bot facing an unknown seller, or a procurement assistant negotiating with a supplier. In such interactions, the counterpart's LLM, prompts, control logic, and rule-based fallbacks are hidden, while each decision can have monetary consequences. We ask whether an agent can predict an unfamiliar counterpart's next decision from a few interactions. To avoid real-world logging confounds, we study this problem in controlled bargaining and negotiation games, formulating it as target-adaptive text-tabular prediction: each decision point is a table row combining structured game state, offer history, and dialogue, while $K$ previous games of the same target agent, i.e., the counterpart being modeled, are provided in the prompt as labeled adaptation examples. Our model is built on a tabular foundation model that represents rows using game-state features and LLM-based text representations, and adds LLM-as-Observer as an additional representation: a small frozen LLM reads the decision-time state and dialogue; its answer is discarded, and its hidden state becomes a decision-oriented feature, making the LLM an encoder rather than a direct few-shot predictor. Training on 13 frontier-LLM agents and testing on 91 held-out scaffolded agents, the full model outperforms direct LLM-as-Predictor prompting and game+text features baselines. Within this tabular model, Observer features contribute beyond the other feature schemes: at $K=16$, they improve response-prediction AUC by about 4 points across both tasks and reduce bargaining offer-prediction error by 14%. These results show that formulating counterpart prediction as a target-adaptive text-tabular task enables effective adaptation, and that hidden LLM representations expose decision-relevant signals that direct prompting does not surface.