Too Consistent to Detect: A Study of Self-Consistent Errors in LLMs
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866908523489656832 |
|---|---|
| author | Tan, Hexiang Sun, Fei Liu, Sha Su, Du Cao, Qi Chen, Xin Wang, Jingang Cai, Xunliang Wang, Yuanzhuo Shen, Huawei Cheng, Xueqi |
| author_facet | Tan, Hexiang Sun, Fei Liu, Sha Su, Du Cao, Qi Chen, Xin Wang, Jingang Cai, Xunliang Wang, Yuanzhuo Shen, Huawei Cheng, Xueqi |
| contents | As large language models (LLMs) often generate plausible but incorrect content, error detection has become increasingly critical to ensure truthfulness. However, existing detection methods often overlook a critical problem we term as self-consistent error, where LLMs repeatedly generate the same incorrect response across multiple stochastic samples. This work formally defines self-consistent errors and evaluates mainstream detection methods on them. Our investigation reveals two key findings: (1) Unlike inconsistent errors, whose frequency diminishes significantly as the LLM scale increases, the frequency of self-consistent errors remains stable or even increases. (2) All four types of detection methods significantly struggle to detect self-consistent errors. These findings reveal critical limitations in current detection methods and underscore the need for improvement. Motivated by the observation that self-consistent errors often differ across LLMs, we propose a simple but effective cross-model probe method that fuses hidden state evidence from an external verifier LLM. Our method significantly enhances performance on self-consistent errors across three LLM families. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_17656 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Too Consistent to Detect: A Study of Self-Consistent Errors in LLMs Tan, Hexiang Sun, Fei Liu, Sha Su, Du Cao, Qi Chen, Xin Wang, Jingang Cai, Xunliang Wang, Yuanzhuo Shen, Huawei Cheng, Xueqi Computation and Language As large language models (LLMs) often generate plausible but incorrect content, error detection has become increasingly critical to ensure truthfulness. However, existing detection methods often overlook a critical problem we term as self-consistent error, where LLMs repeatedly generate the same incorrect response across multiple stochastic samples. This work formally defines self-consistent errors and evaluates mainstream detection methods on them. Our investigation reveals two key findings: (1) Unlike inconsistent errors, whose frequency diminishes significantly as the LLM scale increases, the frequency of self-consistent errors remains stable or even increases. (2) All four types of detection methods significantly struggle to detect self-consistent errors. These findings reveal critical limitations in current detection methods and underscore the need for improvement. Motivated by the observation that self-consistent errors often differ across LLMs, we propose a simple but effective cross-model probe method that fuses hidden state evidence from an external verifier LLM. Our method significantly enhances performance on self-consistent errors across three LLM families. |
| title | Too Consistent to Detect: A Study of Self-Consistent Errors in LLMs |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2505.17656 |