Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | https://arxiv.org/abs/2601.08626 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866909992999714816 |
|---|---|
| author | He, Yingjie Kang, Zhaolu Jiang, Kehan Zhang, Qianyuan Qian, Jiachen Meng, Chunlei Feng, Yujie Wang, Yuan Dou, Jiabao Wu, Aming Zheng, Leqi Zhao, Pengxiang Liu, Jiaxin Zhang, Zeyu Wang, Lei Wang, Guansu Zhan, Qishi He, Xiaomin Zhang, Meisheng Ni, Jianyuan |
| author_facet | He, Yingjie Kang, Zhaolu Jiang, Kehan Zhang, Qianyuan Qian, Jiachen Meng, Chunlei Feng, Yujie Wang, Yuan Dou, Jiabao Wu, Aming Zheng, Leqi Zhao, Pengxiang Liu, Jiaxin Zhang, Zeyu Wang, Lei Wang, Guansu Zhan, Qishi He, Xiaomin Zhang, Meisheng Ni, Jianyuan |
| contents | Large language models (LLMs) excel at semantic understanding, yet their ability to reconstruct internal structure from scrambled inputs remains underexplored. Sentence-level restoration is ill-posed for automated evaluation because multiple valid word orders often exist. We introduce OrderProbe, a deterministic benchmark for structural reconstruction using fixed four-character expressions in Chinese, Japanese, and Korean, which have a unique canonical order and thus support exact-match scoring. We further propose a diagnostic framework that evaluates models beyond recovery accuracy, including semantic fidelity, logical validity, consistency, robustness sensitivity, and information density. Experiments on twelve widely used LLMs show that structural reconstruction remains difficult even for frontier systems: zero-shot recovery frequently falls below 35%. We also observe a consistent dissociation between semantic recall and structural planning, suggesting that structural robustness is not an automatic byproduct of semantic competence. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_08626 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | How Order-Sensitive Are LLMs? OrderProbe for Deterministic Structural Reconstruction He, Yingjie Kang, Zhaolu Jiang, Kehan Zhang, Qianyuan Qian, Jiachen Meng, Chunlei Feng, Yujie Wang, Yuan Dou, Jiabao Wu, Aming Zheng, Leqi Zhao, Pengxiang Liu, Jiaxin Zhang, Zeyu Wang, Lei Wang, Guansu Zhan, Qishi He, Xiaomin Zhang, Meisheng Ni, Jianyuan Computation and Language Large language models (LLMs) excel at semantic understanding, yet their ability to reconstruct internal structure from scrambled inputs remains underexplored. Sentence-level restoration is ill-posed for automated evaluation because multiple valid word orders often exist. We introduce OrderProbe, a deterministic benchmark for structural reconstruction using fixed four-character expressions in Chinese, Japanese, and Korean, which have a unique canonical order and thus support exact-match scoring. We further propose a diagnostic framework that evaluates models beyond recovery accuracy, including semantic fidelity, logical validity, consistency, robustness sensitivity, and information density. Experiments on twelve widely used LLMs show that structural reconstruction remains difficult even for frontier systems: zero-shot recovery frequently falls below 35%. We also observe a consistent dissociation between semantic recall and structural planning, suggesting that structural robustness is not an automatic byproduct of semantic competence. |
| title | How Order-Sensitive Are LLMs? OrderProbe for Deterministic Structural Reconstruction |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2601.08626 |