具有持久化记忆的个性化大语言模型正被越来越多地部署，但其用户模型的忠实度仍未得到检验。我们研究了过度推断（OI）现象：即大语言模型在证据支持范围之外捏造用户属性的行为。我们推出了 MirageBench，包含 150 个在刻板印象、反刻板印象和中性画像之间均衡分布的人物画像，6 项覆盖“想象梯度”的个性化任务，一个由独立评判器（在 400 条声明上与盲法人工标注者进行了验证：四分类 Cohen's kappa = 0.863，二分类 kappa = 0.900）操作的四维忠实度分类体系，以及一个涵盖 7 个模型家族共 12 个模型、基于 143616 条已评判声明的排行榜。
我们发现过度推断普遍存在：全部 12 个模型中有 35%–49% 的声明存在过度推断（跨模型均值 41.6%；按声明加权 41.8%），本次评估中没有任何模型能够幸免。最引人注目的是，我们发现了一种“自我监控反转”现象：在模型选择层面，模型的自我评估 OI 与其评判器测量的 OI 呈负秩相关（rho = -0.60，p = 0.044；探索性结果，宽 bootstrap 置信区间 [-0.90, +0.06]，n = 12）。
自我报告过度推断最少的模型往往被标记为捏造最多的模型，因此自我报告的信心在比较模型时是一个具有误导性的信号——尽管在单个模型内部，自我审计仍能对其自身声明进行中等程度的排序（AUROC 0.58–0.83）。我们进一步表明，OI 具有任务依赖性（27%–59%），并且在多轮试点中，推断出的属性近似线性累积，且很少被修正。
MirageBench 将外部验证而非模型自我报告定位为可信个性化更可靠的基础。
Personalized LLMs with persistent memory are increasingly deployed, yet the faithfulness of their user models remains unexamined. We study over-inference (OI): the phenomenon where LLMs fabricate user attributes beyond what evidence supports. We introduce MirageBench, comprising 150 personas balanced across stereotypical, counter-stereotypical, and neutral profiles, 6 personalization tasks spanning an ``imagination gradient'', a four-way faithfulness taxonomy operationalized by an independent judge (validated against a blind human annotator on 400 claims: Cohen's kappa = 0.863 four-class, kappa = 0.900 binary), and a leaderboard of 12 models across 7 families on 143616 judged claims.
We find that over-inference is pervasive: every one of the 12 models over-infers 35%--49% of its claims (cross-model mean 41.6%; claim-weighted 41.8%), with no model in this evaluation escaping it. Most strikingly, we surface a Self-Monitoring Inversion: at the model-selection level, models' self-assessed OI is negatively rank-correlated with their judge-measured OI (rho = -0.60, p = 0.044; exploratory, wide bootstrap CI [-0.90, +0.06], n = 12). The models that report the least over-inference tend to be flagged as fabricating the most, so self-reported confidence is a misleading signal for comparing models, even though within a single model self-audit still ranks that model's own claims moderately well (AUROC 0.58--0.83).
We further show that OI is task-dependent (27%--59%) and that, in a multi-turn pilot, inferred attributes accumulate approximately linearly with little revision. MirageBench positions external verification, rather than model self-report, as a more reliable foundation for trustworthy personalization.