arXiv cs.CL· Bolei Ma, Yusuke Miyao·· 13 小时前AI 评分31
大语言模型中的未完成体悖论：LLM 存在目的论偏见，将目标导向事件误判为已完成
The Imperfective Paradox in Large Language Models
AI 导读
研究构建 ImperfectiveNLI 诊断数据集，考察大语言模型能否区分进行体在活动动词与达成动词上的蕴含差异。评估发现当前开源权重模型普遍存在目的论偏见，会系统性幻觉出目标导向事件的完成，甚至覆盖文本中明确的取消表述。提示词干预只能部分缓解该偏见，却引发校准危机，使模型错误拒绝无终活动词的有效蕴含。
正文 · 原文
Abstract:Do Large Language Models (LLMs) genuinely grasp the compositional semantics of events, or do they rely on surface-level probabilistic heuristics? We investigate the Imperfective Paradox, a logical phenomenon where the past progressive aspect entails event realization for activities (e.g., running $\to$ ran) but not for accomplishments (e.g., building $\nrightarrow$ built). We introduce ImperfectiveNLI, a diagnostic dataset designed to probe this distinction across diverse semantic classes. Evaluating state-of-the-art open-weight models, we uncover a pervasive Teleological Bias: models systematically hallucinate completion for goal-oriented events, even overriding explicit textual cancellation. Prompting interventions partially reduce this bias but trigger a calibration crisis, causing models to incorrectly reject valid entailments for atelic verbs. Representational analyses further show that while internal embeddings often distinguish progressive from simple past forms, inference decisions are dominated by strong priors about goal attainment. Taken together, our findings indicate that these current open-weight LLMs operate as predictive narrative engines rather than faithful logical reasoners, and that resolving aspectual inference requires moving beyond prompting toward structurally grounded alignment.
| Comments: | ACL 2026. Updates: an additional metric column tbr_a added to table 3; several wording and calculation fixes |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2601.09373 [cs.CL] |
| (or arXiv:2601.09373v3 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2601.09373 arXiv-issued DOI via DataCite |
Submission history
From: Bolei Ma [view email]
[v1]
Wed, 14 Jan 2026 10:57:16 UTC (139 KB)
[v2]
Wed, 22 Apr 2026 11:03:19 UTC (144 KB)
[v3]
Wed, 21 Oct 2026 23:07:50 UTC (144 KB)
来源：arXiv cs.CL · arxiv.org