分离记忆与工作流效应在预测个体答案中的作用
Separating Memory and Workflow Effects in Predicting Individual Answers
浏览论文内容
中文总结 AI 辅助
本研究分离个性化语言智能体的记忆选择与使用方式，提出OwnWords方法，通过BM25检索个人原话进行单次回答，在有限上下文下优于书面记忆，但未超越近期截断，揭示了证据构建过程而非逐字措辞的影响。
中文摘要 AI 辅助
个性化语言智能体既选择记住关于一个人的哪些信息，也选择如何使用这些记忆。在预测已知面试问题的未见答案时，我们将这两种选择分开处理。在来自188人的1,768个任务上，由经过验证的面试前缀构建的具体记忆，其得分比特质描述高出0.0158（95%全人区间[0.0044, 0.0271]）。将两种记忆与一次性生成和三答案融合交叉组合，融合使具体记忆得分降低0.0123（[-0.0189, -0.0056]）；提示式选择器和训练式选择器均未明显优于随机候选。对更长的未重写源记录进行一次调用，其得分超过所有记忆条件。在有限上下文预算下，OwnWords使用BM25检索该人的句子，并在一次调用中给出答案。它在基准之外的500人上优于书面记忆（+0.0127, [+0.0037, +0.0217]；早期留出测试结果不显著），并在300人上跨四个预算平均优于书面记忆（平均+0.0218, [+0.0138, +0.0298]），后者结果在114人上重复。它未明显优于近期截断。这些结果比较了证据构建过程；它们并未分离逐字措辞的影响。在Twin-2K-500上，OwnWords预测序数调查答案比书面记忆更接近，但未提高精确选择准确率，并在两个样本之一中降低了该准确率。面试评分使用基于模型的、无人工评分的、基于内容的评分标准，且原始基准的参与者在开发期间被看到。这些结果描述了所测试过程的特点，而非一般人类预测上限。
英文摘要
Language agents choose what to remember about a person and how to use that memory. We separate these choices when predicting a person's unseen answer to an interview question. On 1,768 tasks from 188 people, a concrete memory from a verified interview prefix outscores a trait description by 0.0158 (95% interval [0.0044, 0.0271]). Crossing both memories with one-shot generation and three-answer fusion, fusion lowers concrete-memory scores by 0.0123 ([-0.0189, -0.0056]); prompted and trained selectors do not detectably beat a random candidate, and one call on the longer, unrewritten record outscores every memory condition. At matched context budgets, OwnWords, one call on the person's BM25-ranked sentences, outperforms the written memory on 500 people outside the benchmark (+0.0127, [+0.0037, +0.0217]; an earlier held-out test was inconclusive) and across four budgets on 300 people (mean +0.0218, [+0.0138, +0.0298]), but does not detectably outperform recency truncation; these comparisons do not isolate verbatim wording. On a survey benchmark it predicts ordinal answers more closely than the memory but is not more accurate on exact choices (less accurate in one of two screener samples). Interview scores use a model-based content rubric without human ratings, and all benchmark and some confirmation participants were seen during development.
发表机构
- University of Cambridge（剑桥大学）
- Stanford University（斯坦福大学）
- Hong Kong Baptist University（香港浸会大学）
- Cookiy Labs（Cookiy实验室）
机构由 AI 辅助整理，请以论文原文为准。
